Optimize lstripChars, rstripChars, and stripChars via specialization

[JoshRosen]

This PR optimizes lstripChars / rstripChars / stripChars built-in functions by using the specialization framework from #119 / 0bd255a to pre-compile and re-use Pattern instances when the replacement / strip arguments are constants.

In Java, String.replaceAll() compiles and uses a Pattern under the hood and this is relatively expensive; specialization lets us save this cost when the chars-to-be-stripped are constant.

Testing

Correctness: ran all tests with a manual change to disable static function application optimizations (a prerequisite required to achieve full test coverage, as the static application prevents specialization from kicking in during most tests). In a followup, I think we should explore adding flags for optimizer features and running all tests with/without optimization.

Performance: ran benchmarks on a large file and with this change I save ~11% of allocated bytes and ~8% of wall time. I ran that same file through RunProfiler and saw a large speedup in stripChars, costing ~619ns/call before and ~154ns/call after.

[JoshRosen]

I considered specializing strReplace but doing so doesn't help (and is actually slower) on JDK 9+ because https://bugs.openjdk.org/browse/JDK-8058779 optimized java.lang.String.replace(CharSequence, CharSequence).