shell/syntax: benchmark and optimize parser

[ewhauser]

Summary

• add an nvm.sh parser benchmark fixture and benchmark coverage under shell/syntax
• reduce parser allocation churn in source-tail handling, raw-word parsing, paren ambiguity probing, and simple-command separator tracking
• add scalar assignment fast paths in callExpr to cut CPU and allocation overhead on common foo=bar cases

Why

The new nvm.sh benchmark made the parser hot spots concrete. Allocation profiles were dominated by parser construction and source slicing, and CPU profiles still showed overhead in separator bookkeeping and assignment-candidate parsing.

Impact

On BenchmarkParseBenchmarkFiles/nvm, the latest controlled benchstat comparison versus the pre-CPU-pass state was:

• sec/op: 7.879 ms -> 5.973 ms (-24.19%)
• B/op: 4.195 MiB -> 4.143 MiB (-1.22%)
• allocs/op: 54.97k -> 46.86k (-14.75%)

Overall from the original baseline to the current parser:

• sec/op: 10.173 ms -> 5.973 ms
• B/op: 19.087 MiB -> 4.143 MiB
• allocs/op: 56.46k -> 46.86k

Validation

• go test ./shell/syntax
• make lint
• env GOMAXPROCS=1 go test ./shell/syntax -run '^$' -bench 'BenchmarkParseBenchmarkFiles/nvm$' -benchmem -cpu 1 -count 15
• benchstat /tmp/gbash-benchstat.7SxhXk/current.txt /tmp/gbash-benchstat.7SxhXk/cpufix.txt

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: d510391ac6

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".