feat: parse escape sequences in unquoted fields (closes #168)

[jsoizo]

Summary

• Make the reader interpret escape sequences (\\, \") in unquoted fields under explicit-escape dialects (escapeChar != quoteChar, e.g. CsvDialect(escapeChar = '\\')), bringing kotlin-csv in line with Python csv, Apache Commons CSV, and OpenCSV.
• The writer-side counterpart (force-quoting fields that contain the escape character) already shipped in a934753; this PR completes the reader side and the "writer is strict / reader is lenient" round-trip story.
• RFC 4180 default (escapeChar == quoteChar == '"') behaviour is unchanged — the helper's {escapeChar, quoteChar} accept set degenerates to a single element so e.g. a"b still throws (covered by existing SequenceParserTest.escapeChar_inFieldState_invalidEscape_throws).

Design notes

Routed all three unquoted states (START, DELIMITER, FIELD) through a single handleUnquotedEscape helper so the accept set and error message are uniform. In START / DELIMITER the quoteChar arm is placed before the escapeChar arm so the escapeChar == quoteChar case is unreachable (handled by QUOTE_START transition). The change was reviewed in plan form by both an internal planning agent and Codex; both independently flagged an earlier draft that special-cased escapeChar == quoteChar inside FIELD as breaking the existing default-dialect throw test — fixed before implementation.

Strict semantics retained: escape followed by EOF / \r / \n / any non-escapeChar-or-quoteChar character throws CsvParseFormatException.

Closes #168.

Test plan

• ./gradlew check green on all multiplatform targets (JVM / JS / Linux / macOS / MinGW / wasmWasi).
• New manual cases in CsvReaderTest: x,a\\b → [["x","a\\b"]], \\b → [["\\b"]], escape + " at START / DELIMITER / FIELD, escape at EOF (throw), escape + ordinary char (throw), escape + CR / LF (throw), multi-row.
• New PBT UnquotedEscapePropertyTest (500 iterations, seed 0L) generates unquoted-safe fields under CsvDialect(escapeChar = '\\'), serialises with Python-csv-style escaping, and round-trips through the reader.
• Verification ritual: intentionally broke handleUnquotedEscape (extra field.append(escapeChar)); both the manual test and PBT went red; reverted.
• Existing escape regression tests in SequenceParserTest (escapeChar_inFieldState_invalidEscape_throws, escapeCharDifferent_invalidEscape_throws, escapeCharDifferent_escapeAtEof_throws, etc.) still pass.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 94.77%. Comparing base (5d062c9) to head (3ab4b2a).
⚠️ Report is 1 commits behind head on version_2_0_0.

Additional details and impacted files

@@                Coverage Diff                @@
##           version_2_0_0     #174      +/-   ##
=================================================
+ Coverage          94.49%   94.77%   +0.27%     
=================================================
  Files                 22       22              
  Lines                418      421       +3     
  Branches              93       95       +2     
=================================================
+ Hits                 395      399       +4     
  Misses                14       14              
+ Partials               9        8       -1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.