This change refactors the final length check in str_casecmp to use existing pointers instead of recalculating the string lengths.

Currently, after the comparison loop finishes, str_casecmp uses RSTRING_LEN to check the lengths of str1 and str2. This requires an additional calculation.

This PR replaces the RSTRING_LEN calls with a check against the p1 and p2 pointers and their respective end pointers (p1end and p2end). Since these pointers are already advanced during the comparison loop, this approach avoids redundant length calculations and slightly improves performance.

The new logic is:

• If both pointers have reached their end (p1 == p1end && p2 == p2end), the strings are equal in length, returning 0.
• If p1 has reached its end but p2 has not (p1 == p1end), str1 is shorter, returning -1.
• Otherwise, p2 has reached its end but p1 has not, so str2 is shorter, returning 1.

Expecting a 3-5% performance increase for str_casecmp.

As I did run benchmark 5 times and took average on my M1 apple;

Benchmarks against master branch:

Benchmarks against current branch: