fix: allow startxref offset to include leading whitespace

[vitormattos]

Summary

• Accept startxref offsets that point to leading whitespace before the xref keyword.
• Use the whitespace-adjusted offset for classic xref decoding and Unix line-ending detection.
• Keep trailer key parsing strict for dictionary entries.

Problem

• Some PDFs set startxref to the whitespace immediately before xref.
• The parser required an exact xref position and incorrectly fell back to xref stream decoding.
• This can lead to Invalid object reference failures for files that are otherwise parseable.

Fix

• Skip PDF whitespace before testing xref position.
• Decode classic xref from the adjusted offset.
• Apply the same adjustment in the Unix line-ending check path.
• Keep strict trailer key matching to avoid accidental value matches.

Regression coverage

• Added fixtures:
  • samples/bugs/PullRequest797-vera.pdf
  • samples/bugs/PullRequest797-pdf.js.pdf
• Added tests:
  • DocumentIssueFocusTest::testParseFileWhenStartxrefPointsToLeadingWhitespaceInVeraPdfFixture
  • DocumentIssueFocusTest::testParseFileWithCompressedXrefObjectFromPdfJsCorpus

Validation

• make run-phpunit ARGS="tests/PHPUnit/Integration/DocumentIssueFocusTest.php"
• Result: OK

PDF Sources

• veraPDF corpus
  • Fixture origin: PDF_A-2b/6.6 Metadata/6.6.2 Metadata streams/6.6.2.3 Schemas/6.6.2.3.2 Extension schemas/veraPDF test suite 6-6-2-3-2-t01-pass-c.pdf
  • Source URL: https://github.com/veraPDF/veraPDF-corpus/blob/staging/PDF_A-2b/6.6%20Metadata/6.6.2%20Metadata%20streams/6.6.2.3%20Schemas/6.6.2.3.2%20Extension%20schemas/veraPDF%20test%20suite%206-6-2-3-2-t01-pass-c.pdf
  • Local copy used in tests: samples/bugs/PullRequest797-vera.pdf
• pdf.js corpus
  • Fixture origin: test/pdfs/pdfkit_compressed.pdf
  • Source URL: https://github.com/mozilla/pdf.js/blob/master/test/pdfs/pdfkit_compressed.pdf
  • Local copy used in tests: samples/bugs/PullRequest797-pdf.js.pdf

[vitormattos]

Added one more regression sample from the pdf.js corpus: (copied as ). It exercises the same startxref/xref recovery path and passes with the current fix.

[vitormattos]

Added one more regression sample from the pdf.js corpus: pdfkit_compressed.pdf (copied as samples/bugs/PullRequest797.pdf). It exercises the same startxref/xref recovery path and passes with the current fix.

[vitormattos]

Renamed the two regression fixtures to make the source corpus explicit: PullRequest797-vera.pdf and PullRequest797-pdf.js.pdf.

[vitormattos]

This PR has been restacked into a per-file consolidation flow for RawDataParser changes.\n\nSuperseded-by chain:\n- base (upstream): #796\n- stacked continuation (fork): https://github.com/vitormattos/pdfparser/pull/26\n\nThe stacked branch keeps equivalent PR797 fix intent while reducing cross-PR conflicts in shared test files. Closing this standalone PR to avoid duplicate merge paths.