fix(dashboard): preserve issue row urls

[corylanou]

Context

Fixes #93. Dashboard issue URL controls need stable row URLs after recent entry-shape changes.

TL;DR

Keep dashboard issue URL buttons wired across running, retry, and recent rows.

Summary

• Normalize orchestrator issue fields through one helper for runtime and retry entries.
• Preserve URL, title, and description when refreshing, blocking, retrying, and snapshotting.
• Add presenter and LiveView regressions for running, retry, and recent URL controls.

Alternatives

• Only patching presenter fallbacks would leave duplicated orchestrator entry shapes fragile.

Test Plan

• make -C elixir all
• mix test test/symphony_elixir/orchestrator_status_test.exs test/symphony_elixir_web/dashboard_live_test.exs
• mix test test/symphony_elixir/extensions_test.exs

[corylanou]

@codex review

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: f398a3621a

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Avoid clearing cached issue fields on running refresh

refresh_running_issue_state/2 now does Map.merge(issue_entry_fields(issue)), which overwrites previously cached :url, :title, and :description with nil when a refreshed %Issue{} is partial. In that case the dashboard loses issue URL controls for still-running rows (and the same pattern is used for blocked rows), whereas the previous behavior preserved the cached top-level fields as a fallback. This shows up when a tracker intermittently omits those fields during poll refreshes.

Useful? React with 👍 / 👎.

[corylanou]

Rework

P2 from Codex on f398a36 — substantively important (re-introduces the original bug):

refresh_running_issue_state/2 now does Map.merge(issue_entry_fields(issue)), which overwrites previously cached :url, :title, and :description with potentially-nil values from the new fetch.

This means the icons LOSE their URL on every refresh cycle — exactly the bug this PR is supposed to lock down. The icons will disappear again after the next state refresh.

Fix: Don't Map.merge with the new fetch values. Instead, only overwrite a cached field if the new value is non-nil. A helper like:

defp merge_keeping_non_nil(existing, new) do
  Enum.reduce(new, existing, fn
    {_k, nil}, acc -> acc
    {k, v}, acc -> Map.put(acc, k, v)
  end)
end

Critical for the regression test: add a test asserting that after a refresh cycle where the fetch returns nil for url, the cached URL is preserved (not nilled). That's the actual lock-in. Without that test, this same bug returns next refactor.