reject overlong decimal integer instead of coercing to inf

[netliomax25-code]

Summary

1. _parse_number runs int(sign + clean, base) inside a try/except ValueError that, on failure, falls back to float(sign + clean).
2. A decimal integer literal with more digits than Python's int-from-string limit (4300 by default) makes int() raise ValueError, so the literal is parsed as a float instead. A long run of digits ends up as inf, so a = 9999... (4301+ digits) silently returns a float inf rather than being rejected as an out-of-range integer.
3. Restrict the float fallback to literals that are actually float-shaped (a fractional dot, a base-10 exponent, or inf/nan). An integer that int() cannot parse now returns None, so the parser raises InvalidNumberError. Hex/oct/bin literals (no digit limit) and all valid float literals are unaffected.

Repro: tomlkit.parse("a = " + "9" * 4301)["a"] returns float inf before the change; it raises InvalidNumberError after. The 4300-digit value still parses as an Integer.

Agent Drafting Metadata

• Agent:
• Model:
• Notes: