[C++][Compute] Cache substring and regex matchers across invocations in scalar string kernels #53159

Describe the enhancement requested

description

in cpp/src/arrow/compute/kernels/scalar_string_ascii.cc, the MatchSubstring kernels (match_substring, starts_with, ends_with, match_substring_regex and match_like) rebuild their matcher on every call to Exec.

for regex matchers that means the RE2 DFA gets recompiled for every batch or chunk in a stream or chunked array. for plain matchers it reallocates and rebuilds the Knuth-Morris-Pratt prefix table each batch.

both MatchSubstring<Type, Matcher>::Exec and MatchSubstring<Type, RegexSubstringMatcher>::Exec already carry a maintainer comment about this:
// TODO Cache matcher across invocations (for regex compilation)

proposed solution

replace OptionsWrapper<MatchSubstringOptions> with a dedicated MatchSubstringState. it lazily builds the matcher once and keeps the compiled or precomputed result in the KernelState, so later invocations reuse it.

Component(s)

C++