perf: refactor to avoid excessive walking

[anderseknert]

Store result of vars found in walk into an intermediate object, keyed by rule index and the variable's "context".

The result is quite an improvement:

❯ hyperfine --warmup 2 'regal_main lint bundle' 'regal_new lint bundle'
Benchmark 1: regal_main lint bundle
  Time (mean ± σ):      4.387 s ±  0.174 s    [User: 31.888 s, System: 0.906 s]
  Range (min … max):    4.260 s …  4.820 s    10 runs

Benchmark 2: regal_new lint bundle
  Time (mean ± σ):      3.502 s ±  0.120 s    [User: 24.250 s, System: 0.688 s]
  Range (min … max):    3.308 s …  3.700 s    10 runs

Summary
  regal_new lint bundle ran
    1.25 ± 0.07 times faster than regal_main lint bundle

[srenatus]

Nicely done ✨