perf(core-go): zero-allocation memo normalization and routing extraction

[TheWeirdDee]

Resolves #265

Replace regexp.MustCompile and big.Int with allocation-free alternatives on the hot path in the Go routing package:

• memo.go: remove digitsOnly regex and uint64Max big.Int package-level vars; add isAllDigits (byte-scan loop) and fitsUint64 (string length + lexicographic compare against uint64MaxStr constant). NormalizeMemoTextID now returns 0 allocs/op on canonical inputs (proved by benchmarks).

• extract.go: remove duplicate digitsOnlyRegex var; rewrite normalizeUnsupportedMemoType with strings.Builder to avoid three intermediate string allocations from chained ReplaceAll/ToLower; use pre-computed string literals for known memo-type warning messages; use strings.Builder for the dynamic unrecognized-type message; pre-allocate warning slices with make([]Warning, 0, cap) instead of appending into a zero-capacity slice.

• bench_test.go: add 11 benchmarks (BenchmarkNormalizeMemoTextID_, BenchmarkIsAllDigits, BenchmarkFitsUint64, BenchmarkExtractRouting_, parallel variants) with b.ReportAllocs(). Results on i7-10510U:
  NormalizeMemoTextID_Canonical 12.7 ns/op 0 allocs/op
  NormalizeMemoTextID_MaxUint64 18.3 ns/op 0 allocs/op
  IsAllDigits 11.3 ns/op 0 allocs/op
  FitsUint64 0.27 ns/op 0 allocs/op
  ExtractRouting_GAddr_MemoID 757 ns/op 6 allocs/op (parse-dominated)
  ExtractRouting_GAddr_Parallel 322 ns/op 6 allocs/op

[drips-wave]

@TheWeirdDee Great news! 🎉 Based on an automated assessment of this PR, the linked Wave issue(s) no longer count against your application limits.

You can now already apply to more issues while waiting for a review of this PR. Keep up the great work! 🚀

Learn more about application limits