TOML: compact integer underscores without a second char array

[skrcode]

This is a one-commit JAIPilot Cloud follow-up on the exact current head of FasterXML/jackson-dataformats-text#713. It targets the original contributor branch so the change can be accepted without a duplicate upstream feature PR.

Change

Compact underscores in place in TomlParser.parseInt using the parser-owned token buffer. Characters before the first underscore are already in place, so the slow path no longer allocates and fills a second char array.

Proof

• Exact parent: 6ce8d82, the current Misc perf fixes (toml, yaml) FasterXML/jackson-dataformats-text#713 head.
• The same 113 focused TOML parser tests passed before and after, including decimal, hex, octal, binary, long-boundary, buffer-growth, and following-token cases.
• Five fixed 1,000-call allocation batches on a 10,000-digit underscored integer: median allocated bytes per parse fell from 551,104 to 524,416, exactly one char array removed.
• Full ./mvnw clean verify passed 1,558 tests with zero failures or errors.

Boundary and disclosure

Only the underscore-bearing slow path changes; integer tokens without underscores are untouched. The allocation fixture forces a very long token and is not a production-throughput claim.

JAIPilot Cloud generated and validated the patch in skrcode/jackson-dataformats-text#2. I reviewed the exact ancestry, complete two-file diff, verification evidence, and limitation before offering it here.