MINOR: Fix VariantUtil.getByte() Javadoc copied from getShort()

[anxkhn]

Rationale for this change

The Javadoc on the package-private helper VariantUtil.getByte(ByteBuffer) is a
verbatim copy of the getShort(ByteBuffer) block directly above it, and was never
adjusted for getByte. It currently reads:

/**
 * Similar to getLong(), but for the types: Type.BYTE, SHORT.
 * @param value The Variant value
 * @return The short value
 */
static byte getByte(ByteBuffer value) {

Both lines contradict the method:

• getByte does not handle Type.SHORT. Its switch has only case INT8: (the
  BYTE type); every other type, SHORT included, falls through to
  throw unexpectedType(Variant.Type.BYTE, value).
• @return The short value contradicts the declared static byte getByte(...)
  return type; the method returns (byte) readLong(value, value.position() + 1, 1).

getShort above it genuinely handles both INT8 and INT16 and returns short,
so the doc block was clearly copy-pasted onto getByte and left stale.

What changes are included in this PR?

A two-line Javadoc correction on getByte (in
parquet-variant/src/main/java/org/apache/parquet/variant/VariantUtil.java):

   /**
-   * Similar to getLong(), but for the types: Type.BYTE, SHORT.
+   * Similar to getLong(), but for the type: Type.BYTE.
    * @param value The Variant value
-   * @return The short value
+   * @return The byte value
    */
   static byte getByte(ByteBuffer value) {

The (correct) getShort Javadoc directly above is left untouched. No code behavior
changes.

Are these changes tested?

No new test. This is a Javadoc correction with no runtime behavior change, so a
unit test would not assert anything meaningful. The corrected wording simply makes
the internal contract match the already-tested behavior (TestVariantScalar
exercises getByte on BYTE values and the invalid-read paths). I confirmed the
module still passes its gates locally:

• mvn -pl parquet-variant spotless:check -> BUILD SUCCESS (palantir-java-format
  gate passes)
• mvn -pl parquet-variant -am -Dspotless.check.skip=true test -> BUILD SUCCESS
  (parquet-variant: 180 run, 0 failures, 0 errors, 1 skipped)

Are there any user-facing changes?

No. Documentation (Javadoc) only, on a package-private helper. No API or behavior
change.