Back parser state exportedIdentifiers by set

[JLHwung]

Q	A
Fixed Issues?	Parsing named exports costs O(n^2) time
License	MIT

This PR changes the data structure of exportedIdentifiers to a set to avoid O(n^2) time when checking duplicate exports. This PR also moves exportedIdentifiers to the root state so it won't be cloned in state.clone.

Benchmark result:

Length-1 named exports

$ node --predictable ./many-named-export/1-length.bench.mjs

baseline 256 length-1 named export: 2060 ops/sec ±20.93% (0.485ms)
baseline 512 length-1 named export: 875 ops/sec ±1.16% (1.143ms)
baseline 1024 length-1 named export: 273 ops/sec ±1.3% (3.669ms)
baseline 2048 length-1 named export: 73.69 ops/sec ±2.17% (14ms)

current 256 length-1 named export: 2862 ops/sec ±34.06% (0.349ms)
current 512 length-1 named export: 1747 ops/sec ±1.39% (0.572ms)
current 1024 length-1 named export: 814 ops/sec ±0.63% (1.228ms)
current 2048 length-1 named export: 392 ops/sec ±0.45% (2.549ms)

[codesandbox-ci]

This pull request is automatically built and testable in CodeSandbox.

To see build info of the built libraries, click here or the icon next to each commit SHA.

Latest deployment of this branch, based on commit 38ea497:

Sandbox	Source
babel-repl-custom-plugin	Configuration
babel-plugin-multi-config	Configuration

[babel-bot]

Build successful! You can test your changes in the REPL here: https://babeljs.io/repl/build/46592/