Previously the inner request object of the original and the new request
were the same, causing the requests to be entangled and mutable changes
to one to be visible to the other. This fixes that.
Avoid "blob:" prefix check on requests built in the http module since those can never be blob objects
Reduces cost of `newInnerRequest()` from 20ms to 0.1ms in my profiled run on ~2.5M reqs
* perf(ext/fetch): skip USVString webidl conv on string constructor
* Rename webidl convert to RequestInfo_DOMString
To disambiguate and hint that it normalizes to DOMString instead of USVString since DOMString => USVString is handled by `op_url_parse` when calling `new URL(...)`
In the spec, a URL record has an associated "blob URL entry", which for
`blob:` URLs is populated during parsing to contain a reference to the
`Blob` object that backs that object URL. It is this blob URL entry that
the `fetch` API uses to resolve an object URL.
Therefore, since the `Request` constructor parses URL inputs, it will
have an associated blob URL entry which will be used when fetching, even
if the object URL has been revoked since the construction of the
`Request` object. (The `Request` constructor takes the URL as a string
and parses it, so the object URL must be live at the time it is called.)
This PR adds a new `blobFromObjectUrl` JS function (backed by a new
`op_blob_from_object_url` op) that, if the URL is a valid object URL,
returns a new `Blob` object whose parts are references to the same Rust
`BlobPart`s used by the original `Blob` object. It uses this function to
add a new `blobUrlEntry` field to inner requests, which will be `null`
or such a `Blob`, and then uses `Blob.prototype.stream()` as the
response's body. As a result of this, the `blob:` URL resolution from
`op_fetch` is now useless, and has been removed.
