Disambiguate duplicate enum variant names instead of panicking

[danieleades]

Problem

Schemas whose enum values sanitize to identical Rust identifiers cause a panic during generation:

panic: Failed to make unique variant names for [=,>,<,≥,>=,≤,<=,≠,!=]

Variant naming makes two attempts (plain sanitization, then replacing non-identifier characters with X), and if names still collide — as with the comparison-operator enum above, where =, >, <, ≥, ≤, and ≠ all reduce to X — it panics.

Fixes #948

Fix

Replace the panic with a final, deterministic disambiguation pass: proceeding in variant order, the first variant with a given name keeps it unchanged, and each subsequent duplicate gets the smallest numeric suffix (starting at 2) that doesn't collide with any other variant name (guarding against, say, a pre-existing literal X2 variant). Three variants that all sanitize to X become X, X2, X3.

Serialization is unaffected: serde renames and the Display/FromStr impls are all generated from the variant's raw name, so suffixed variants round-trip through the original strings. The confusing duplicate-counting bookkeeping that fed the panic message is removed along the way.

Trade-offs

Schemas that previously generated successfully produce identical output; the new pass only runs where generation used to panic, so the suffixed names introduce no compatibility concern.

Tests

• Unit test converting the exact value set from Failed to make unique variant names for [=,>,<,≥,>=,≤,<=,≠,!=] #948, asserting raw names are preserved in order and identifier names are disambiguated as expected.
• New golden pair typify/tests/schemas/colliding-variant-names.{json,rs} covering the same schema end to end (including trybuild compilation).