Performance + Runtime Simplification: keyed r-for diff as default, binder/directive cleanup, benchmark/docs upgrades

[koculu]

Major internal performance/refactor pass:

• Makes optimized r-for paths first-class (ForBinderFastPath + keyed diff).
• Simplifies binder/directive internals and expression binding flow while keeping runtime behavior intact.
• Expands benchmark tooling/UI (DOM + minidom) and profiling workflow (perf:profile) for repeatable perf validation.
• Improves docs broadly, including performance guide, lifecycle cleanup docs, and a new API page for drainUnbind.
• Adds/updates extensive tests for binder/directive reactivity, edge cases, and regression coverage.

Net: cleaner internals, better perf instrumentation, and clearer docs with no intended feature loss.