fix(fsst): adaptive i32→i64 codes_offsets promotion

[mprammer]

Fixes #7833. fsst_compress_iter previously hardcoded VarBinBuilder::<i32> for the FSST output and panicked once cumulative compressed bytes crossed i32::MAX. The companion change to vortex-array/src/arrays/varbin/compute/compare.rs handles the latent Arrow LargeBinary == Binary mismatch that becomes reachable once any FSST output uses i64 offsets.

The codes-offsets buffer now starts as i32 and promotes to i64 on the first append where data.len() crosses i32::MAX — one O(n) copy of existing offsets, sub-millisecond at any realistic n. Common case is identical to the original i32 path. Tests: a small-input round-trip asserts the i32 baseline; a CI-gated ~4.5 GiB test exercises the i64 path end-to-end.

🤖 Generated with Claude Code

[codspeed-hq]

Merging this PR will degrade performance by 24.95%

⚠️ Unknown Walltime execution environment detected

Using the Walltime instrument on standard Hosted Runners will lead to inconsistent data.

For the most accurate results, we recommend using CodSpeed Macro Runners: bare-metal machines fine-tuned for performance measurement consistency.

⚠️ Different runtime environments detected

Some benchmarks with significant performance changes were compared across different runtime environments,
which may affect the accuracy of the results.

Open the report in CodSpeed to investigate

⚡ 2 improved benchmarks
❌ 6 regressed benchmarks
✅ 1200 untouched benchmarks

⚠️ Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
⚡	Simulation	take_map[(0.1, 0.5)]	1,111.5 µs	988.1 µs	+12.49%
⚡	Simulation	take_map[(0.1, 1.0)]	1.8 ms	1.6 ms	+11.89%
❌	Simulation	new_bp_prim_test_between[i32, 16384]	94.9 µs	110 µs	-13.73%
❌	Simulation	new_bp_prim_test_between[i32, 32768]	141.2 µs	170.8 µs	-17.31%
❌	Simulation	new_bp_prim_test_between[i64, 16384]	115.5 µs	145.4 µs	-20.61%
❌	Simulation	new_bp_prim_test_between[i64, 32768]	178.3 µs	237.6 µs	-24.95%
❌	Simulation	new_bp_prim_test_between[i16, 32768]	120.6 µs	135.1 µs	-10.71%
❌	Simulation	new_alp_prim_test_between[f64, 16384]	127.3 µs	149.8 µs	-15.04%

---

_{Comparing mp/fsst-i32-overflow-adaptive (9e2bcdc) with develop (ff12040)}