Perf: Optimize substring_index via single-byte fast path and direct indexing

[lyne7-sc]

Which issue does this PR close?

Rationale for this change

This PR improves the performance of the substring_index function by optimizing
delimiter search and substring extraction:

• Single-byte fast path: introduces a specialized byte-based search for single-byte delimiters (e.g. ., ,), avoiding UTF-8 pattern matching overhead.
• Efficient index discovery: replaces the split-and-sum-length approach with direct index location using match_indices / rmatch_indices.

What changes are included in this PR?

• Added a fast path for delimiter.len() == 1 using byte-based search.
• Refactored the general path to use match_indices and rmatch_indices for more efficient positioning.

Benchmarks

• Single-byte delimiter benchmarks show ~2–3× speedup across batch sizes.
• Multi-byte delimiters see a consistent ~10–15% improvement.

group                                               main_substrindex                       perf_substrindex
-----                                               ----------------                       ----------------
substr_index/substr_index_10000_long_delimiter      1.12   548.2±18.39µs        ? ?/sec    1.00   488.4±15.62µs        ? ?/sec
substr_index/substr_index_10000_single_delimiter    2.14   543.4±15.12µs        ? ?/sec    1.00    254.0±7.80µs        ? ?/sec
substr_index/substr_index_1000_long_delimiter       1.12     43.1±1.63µs        ? ?/sec    1.00     38.6±2.03µs        ? ?/sec
substr_index/substr_index_1000_single_delimiter     3.51     46.4±2.21µs        ? ?/sec    1.00     13.2±0.99µs        ? ?/sec
substr_index/substr_index_100_long_delimiter        1.01      3.7±0.18µs        ? ?/sec    1.00      3.7±0.20µs        ? ?/sec
substr_index/substr_index_100_single_delimiter      2.15      3.6±0.14µs        ? ?/sec    1.00  1675.9±79.15ns        ? ?/sec

Are these changes tested?

• Yes, Existing unit tests pass.
• New benchmarks added to verify performance improvement.

Are there any user-facing changes?

No.

[uzqw]

Hi! I noticed some Clippy warnings when running locally:
cargo clippy -p datafusion-functions --all-targets --all-features

[uzqw]

this else { if .. } block can be collapsed

[uzqw]

variables can be used directly in the format! string

[uzqw]

variables can be used directly in the format! string