refactor(load): unify legacy and opossum file loading paths

[abraemer]

Summary

After the fflate-to-adm-zip migration (#3164), the loading code still branched on file format throughout the entire stack. Every layer — from listeners through the DB process to loadFile.ts — had isOpossumFileFormat checks that selected different parsing paths, different save targets, and different state fields depending on whether the input was a .opossum file or a legacy JSON input.

This PR eliminates that dual-path complexity by ensuring every load produces the same uniform result — a LoadedArchive containing an AdmZip instance, parsed input, and parsed output — regardless of whether the source was a .opossum file or a legacy JSON input. It also simplifies the save path so that loadFile passes the output data it already has in hand directly to writeOpossumFile, avoiding a pointless round-trip through the DB.

Stacked on top of #3164 (adm-zip migration).

Problem

Before this PR, the import flow for legacy files was:

1. importFileConvertAndLoadListener → convertToOpossum() writes a .opossum file to disk
2. initializeGlobalBackendState(opossumFilePath, true) → opens it as .opossum
3. loadFile() → re-reads the file it just wrote, branches on isOpossumFileFormat()
4. Downstream code branches on format for save paths, state fields, etc.

This was wasteful (write + immediate re-read) and needlessly complex: the listener already knew the format but threw that information away, forcing loadFile to re-derive it from the file extension.

Meanwhile, the "direct legacy load" path in loadFile.ts (the else branch of isOpossumFileFormat) was unreachable from the UI and broken — saving would throw Cannot save: no input file loaded because storedOpossumZip was undefined.

Additionally, saveFile covertly depended on global DB state via getSaveFileArgs(), even in the initial load path where the output data was already in hand — forcing a pointless DB round-trip.

Approach

Unified loading

Two format-specific loader functions produce a uniform LoadedArchive type:

• loadOpossumFile(path) — opens AdmZip, extracts and parses input.json/output.json
• loadLegacyFile(path) — reads raw JSON (.json/.json.gz), parses it, reads sidecar _attributions.json if present, then constructs a new in-memory AdmZip with both entries packed in

From that point on, all downstream code is format-agnostic: always an AdmZip, always an opossumFilePath for saves.

The listener layer explicitly picks the right loader (it already knows the format), so loadFile() no longer needs to infer it.

Explicit save paths

loadFile now adds output.json to the in-memory zip when it creates output from scratch (the null-output case), then passes resolvedOutputData directly to writeOpossumFile — no round-trip through the DB. The user-initiated save path in dbProcess.ts explicitly calls buildOpossumOutputFile (which reads from the DB) then writeOpossumFile. The global-state dependency is visible at each call site, not hidden inside a wrapper function.

Changes

Core loading

• parseFile.ts: Replace parseOpossumFile / parseInputJsonFile / parseOutputJsonFile with loadOpossumFile / loadLegacyFile returning uniform LoadedArchive
• loadFile.ts: Remove all isOpossumFileFormat branching; accepts LoadedArchive + opossumFilePath. When creating output from scratch, adds output.json to the zip and writes via writeOpossumFile directly (no DB round-trip)

IPC layer

• dbProcess.ts: Split LoadFileMessage into LoadOpossumFileMessage (format: "opossum") and LoadLegacyFileMessage (format: "legacy"); routes to correct loader, passes uniform result to loadFile. Save handler calls buildOpossumOutputFile + writeOpossumFile explicitly
• dbProcessClient.ts: Split loadFile() into loadOpossumFile() and loadLegacyFile(); saveFile() takes explicit (projectId, opossumFilePath) instead of SaveFileParams
• importFromFile.ts: Split into loadOpossumFileFromPath() and loadLegacyFileFromPath()

Listeners

• listeners.ts: importFileConvertAndLoadListener no longer calls convertToOpossum() (disk write + re-read) for legacy files; calls loadLegacyFileFromPath() directly. ScanCode/OWASP still use convertToOpossum. saveFileListener simplified — no more attributionFilePath/inputFileChecksum. initializeGlobalBackendState simplified — no isOpossumFormat boolean.

Save path

• Deleted saveFile.ts — SaveFileParams and persistOutputFile removed; callers use writeOpossumFile directly
• buildOpossumOutputFile.ts (renamed from getSaveFileArgs.ts): exports buildOpossumOutputFile (assembles OpossumOutputFile from DB state) and getPreferredOverOriginIds (for its own test); querySaveFileArgs is private
• writeOpossumFile in write-file.ts: always uses zip.updateFile(OUTPUT_FILE_NAME, ...) — the entry is guaranteed to exist because loadFile adds it when creating output

Types & state

• types.ts: Remove resourceFilePath/attributionFilePath from GlobalBackendState; remove ParsedOpossumInputAndOutput (replaced by LoadedArchive)
• getLoadedFile.ts: Remove getLoadedFileType() (dead code); simplify getLoadedFilePath() and isFileLoaded()

Deleted files

• enums.ts (only contained LoadedFileFormat enum)
• isOpossumFileFormat.ts
• opossumArchive.ts (already inlined in previous commit)
• saveFile.ts (replaced by direct writeOpossumFile calls)

Test plan

• parseFile.test.ts — 12 tests pass (updated to test loadOpossumFile/loadLegacyFile; added tests for sidecar output reading, .json.gz, error handling on sidecar)
• loadFile.test.ts — updated to use new two-step flow (load archive → loadFile); tests that output.json is written correctly for legacy input
• buildOpossumOutputFile.test.ts — simplified (only tests buildOpossumOutputFile, not the deleted wrapper)
• getLoadedFile.test.ts — simplified (removed LoadedFileFormat.Json tests)
• globalBackendState.test.ts — updated (removed resourceFilePath/attributionFilePath)
• listeners.test.ts / get-save-file-listener.test.ts / importFromFile.test.ts / errorHandling.test.ts — all pass with updated mocks
• opossum-file.test.ts — updated to use loadOpossumFile instead of parseOpossumFile
• TypeScript type check passes (tsc --noEmit)
• Knip (unused export check) passes
• Manual test: open a .opossum file
• Manual test: import a legacy .json file via the import dialog