Improves boot time

[arcanis]

What's the problem this PR addresses?

Yarn is currently a little slow to start. Even a yarn --version takes ~280ms on my (slow) machine.

How did you fix it?

This diff implements bundle splitting by adding a --split option to the bundle generator that, when used, causes it to emit a self-extracting file that unpacks the chunks on disk then runs them. In practice it isn't more performant, so it's unlikely to be enabled by default, but having this capability can be useful for experimentation purposes.

A few other runtime speed improvements not directly related to it are also implemented.

• Reimplemented Tinylogic using Pegjs rather than Chevrotine
• Changed some of the got import into type imports
• Removed yup from the modules exposed to plugins (was legacy)

YARN_IGNORE_PATH=1 hyperfine -w 1 "node ./master.cjs --version" "node ./after.cjs --version"
Benchmark 1: node ./master.cjs --version
  Time (mean ± σ):     214.8 ms ±   1.1 ms    [User: 228.5 ms, System: 18.1 ms]
  Range (min … max):   213.4 ms … 216.9 ms    13 runs

Benchmark 2: node ./after.cjs --version
  Time (mean ± σ):     178.0 ms ±   1.3 ms    [User: 189.2 ms, System: 20.0 ms]
  Range (min … max):   175.7 ms … 180.8 ms    16 runs

Summary
  'node ./after.cjs --version' ran
    1.21 ± 0.01 times faster than 'node ./master.cjs --version'

Checklist

• I have read the Contributing Guide.

• I have set the packages that need to be released for my changes to be effective.

• I will check that all automated PR checks pass before the PR gets reviewed.

[merceyz]

Moved the builtin patch behind lazy imports
Moved the worker hook behind a lazy import
Moved the PnP hook behind a lazy import

Without bundle splitting these can be reverted as they are just functions

[merceyz]

I updated the benchmark results to the current state of this PR