Review + consider transferring to the rdfjs namespace

## What this is

`jsonld-fast-parse` is a specialized, **synchronous** JSON-LD → RDF/JS quad
converter for **fixed-shape, known-context** documents — primarily the
Components.js-generated component files that dominate Community Solid Server /
Comunica boot. It walks a `JSON.parse`d object once against a **pre-normalized**
context and emits RDF/JS quads directly, skipping streaming, async handler
dispatch, and per-file context re-derivation.

**Public API (one line):** `fastParse(document, normalizedContext, options?) => Quad[]`
plus a `canFastParse(text) => boolean` shape guard (and the `Converter` class,
`canonicalJson`, and `UNSUPPORTED_PATTERN` for advanced use).

## Performance evidence

Measured on the CSS boot corpus — **828** Components.js `*.jsonld` files, warmed
process, best-of-N `process.hrtime`, Node 22:

| path | best time | per file |
| --- | --- | --- |
| generic `rdf-parse` pipeline (Comunica JSON-LD → `jsonld-streaming-parser`, 8-way concurrent, prefetched doc loader) | ~6050 ms | ~7.3 ms |
| specialized `JSON.parse` + `fastParse` (contexts pre-normalized) | ~19 ms | ~23 µs |

- **~279–316× steady-state speedup** (varies with host load); ~35× including the
one-time ~155 ms normalization of all 169 distinct contexts; ~2.4× overhead
over a bare `JSON.parse` of the same files.
- **Parity is verified, not assumed:** identical quad count (34 926) and, for
**every** file, the two quad sets are **`rdf-isomorphic`** (identical modulo
blank-node labels) — the same 8 504-resource object graph.

## Correctness contract (important)

This is **NOT a general-purpose JSON-LD processor**. It implements only a
documented subset — `@id`, `@type`, `@value` (+ `@type`/`@language`/`@json`),
`@list` (`@container: @list` and inline), `@type: "@id"` coercion, compact IRIs /
prefixes, native scalars with spec-canonical lexical forms.

It explicitly does **not** support `@graph`, `@reverse`, `@index`, `@nest`,
`@included`, `@direction`, `@base`, `@vocab`, type-scoped / nested inline
contexts, framing, language maps, or non-`@list` containers, and it requires the
context to be pre-normalized. Callers **must** guard with `canFastParse(text)`
(which rejects the detectable unsupported keywords) and fall back to a
spec-compliant parser otherwise. The README leads with this contract, and the
test suite (`npm test`) reproduces the `rdf-isomorphism` oracle against
`jsonld-streaming-parser` on bundled sample `.jsonld` fixtures. `npm install`,
`npm run build` (tsc), and `npm test` all pass; runtime deps are limited to
`@rdfjs/types` + `rdf-data-factory`.

## Provenance

This repository was drafted by @jeswr's AI agent at his request, as an initial
0.1.0 packaging of a performance prototype. Nothing has been published to npm —
it is a GitHub repo only.

## Ask

@jeswr — please review the code, README, and the correctness contract / guard,
and sanity-check the benchmark methodology and the supported/unsupported subset.

Once you're happy with it, I'd like to **offer to transfer this repository to the
[`rdfjs`](https://github.com/rdfjs) organisation** so it can live alongside the
other RDF/JS tooling. You can do that yourself via **Settings → General →
Danger Zone → Transfer ownership** (transfer to `rdfjs`), or by asking an
`rdfjs` org admin to accept the transfer. Happy to adjust naming, scope, or the
API before any transfer if you'd prefer.