docs: repair seven broken javadoc references

`javadoc:javadoc` failed with "reference not found" in three modules.
Every case was a stale link, mostly left by past refactors:

- StructArray: [io.github.dfa1.vortex.encoding.StructEncoding] — that
  package no longer exists; the type is now
  reader.decode.StructEncodingDecoder.
- NullableData: [Encoding] — split into EncodingEncoder/EncodingDecoder.
- ArrayStats: [#mostFrequentValueBits()] — the record component is
  mostFrequentBits.
- Chunk: [DType] and ByteBoolEncodingDecoder: [BoolArray] — neither type
  is imported, so neither resolved. Chunk now uses the FQN (DType appears
  only in prose); ByteBool points at MaterializedBoolArray, which is both
  imported and what decode() actually returns.
- NullableData: [io.github.dfa1.vortex.reader.array.MaskedArray] and
  CodeGen: [FbsTable]/[FbsBuilder] — cross-module targets that are
  deliberately unreachable. Writer must not depend on reader, and fbs-gen
  emits the core.fbs runtime classes rather than depending on them, so
  both are now named as text with the reason stated.

Comment-only: no signature, no behavior. Verified with
`./mvnw package -DskipTests javadoc:javadoc -fae`, which is now clean
across all 17 modules; checkstyle passes.

Note that bare `./mvnw javadoc:javadoc` still fails, but on dependency
resolution rather than javadoc: the goal runs no lifecycle phase, so the
reactor jars do not exist, and the project forbids `mvn install`. Prefix
it with `package -DskipTests` to run the check.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

Author: dfa1

fbs-gen/src/main/java/io/github/dfa1/vortex/fbsgen/CodeGen.java

@@ -16,7 +16,10 @@
 /// - each `enum` becomes a constants holder over its underlying integer type;
 /// - each `union` becomes a discriminator-constants holder (NONE = 0, members from 1).
 ///
-/// The wire layout is the standard FlatBuffers binary format — see [FbsTable]/[FbsBuilder].
+/// The wire layout is the standard FlatBuffers binary format. The runtime base classes the
+/// emitted code extends (`FbsTable`, `FbsMemorySegment`, `FbsBuilder`) live in the generated
+/// package `io.github.dfa1.vortex.core.fbs`, which this module does not depend on — it emits
+/// their source, so they are named here as text rather than linked.
 public final class CodeGen {
 
     /// Prefix applied to every emitted Java class name so the generated wire-format

reader/src/main/java/io/github/dfa1/vortex/reader/array/StructArray.java

@@ -10,7 +10,7 @@
 
 /// Decoded struct array: holds one [Array] per field, keyed by position.
 ///
-/// Returned by [io.github.dfa1.vortex.encoding.StructEncoding] when decoding a
+/// Returned by [io.github.dfa1.vortex.reader.decode.StructEncodingDecoder] when decoding a
 /// multi-field `vortex.struct` segment (all columns packed into one flat).
 public final class StructArray implements Array {
 

reader/src/main/java/io/github/dfa1/vortex/reader/decode/ByteBoolEncodingDecoder.java

@@ -8,7 +8,7 @@
 import java.lang.foreign.ValueLayout;
 
 /// Read-only decoder for `vortex.bytebool` — packs the input byte buffer into the
-/// bit-packed [BoolArray] layout used by `vortex.bool`.
+/// bit-packed [MaterializedBoolArray] layout used by `vortex.bool`.
 public final class ByteBoolEncodingDecoder implements EncodingDecoder {
 
     @Override

writer/src/main/java/io/github/dfa1/vortex/writer/Chunk.java

@@ -6,7 +6,7 @@
 ///
 /// Validates each `put` against the writer's schema:
 /// - Column name must exist in the schema.
-/// - Array type must match the column [DType].
+/// - Array type must match the column [io.github.dfa1.vortex.core.model.DType].
 /// - Non-nullable columns reject boxed arrays containing `null`.
 ///
 /// All schema columns must be supplied before the consumer returns; all column

writer/src/main/java/io/github/dfa1/vortex/writer/encode/ArrayStats.java

@@ -11,7 +11,7 @@
 /// sample-encoding probe that biases on leading rows.
 ///
 /// All values are stored as raw 64-bit patterns (`doubleToRawLongBits` for floats,
-/// zero-extended unsigned ints for U8/U16/U32). Use [#mostFrequentValueBits()] etc. and
+/// zero-extended unsigned ints for U8/U16/U32). Use [#mostFrequentBits()] etc. and
 /// reinterpret per ptype as needed.
 ///
 /// @param valueCount        number of non-null elements scanned

writer/src/main/java/io/github/dfa1/vortex/writer/encode/NullableData.java

@@ -3,15 +3,16 @@
 /// Writer-side carrier for nullable column data: a packed values array paired with a
 /// per-row validity bitmap.
 ///
-/// `values` carries the raw storage in the shape an [Encoding] expects
+/// `values` carries the raw storage in the shape an [EncodingEncoder] expects
 /// (`int[]`, `long[]`, `byte[]`, `double[]`, ...). Null
 /// positions hold zero-valued placeholders so primitive encoders can compress them
 /// alongside real data. `validity` has the same logical length: `true`
 /// for valid rows, `false` for nulls.
 ///
 /// The writer recognizes this shape and emits the `vortex.masked`
 /// wire layout: a non-nullable child (the storage) plus an optional Bool
-/// validity child. Readers reconstruct a [io.github.dfa1.vortex.reader.array.MaskedArray].
+/// validity child. Readers reconstruct a `MaskedArray` — named as text, not linked, because
+/// the writer module does not depend on the reader.
 ///
 /// Invariant: `values` length (or storage row count) matches
 /// `validity.length`; placeholders at null positions must be zero so the