responses with transfer-encoding break the urllib handler

[machineonamission]

So, yall construct a fake full HTTP response here

pyodide-http/pyodide_http/_urllib.py

Lines 38 to 50 in b5fc794

	headers_without_content_length = {
	k: v for k, v in resp.headers.items() if k != "content-length"
	}
	response_data = (
	b"HTTP/1.1 "
	+ str(resp.status_code).encode("ascii")
	+ b"\n"
	+ "\n".join(
	f"{key}: {value}" for key, value in headers_without_content_length.items()
	).encode("ascii")
	+ b"\n\n"
	+ resp.body
	)

now, some http requests have transfer-encoding, where the data is chunked or zipped or something. problem is, XMLHttpRequest abstracts that away, but leaves the response headers. So JS passes decoded data, but says its chunked, which freaks out python's http lib.

the fix is simple, strip the transfer-encoding header before constructing the response.

this is confirmed working with chunked transferring. i am not 100% sure of gzip and such, but i imagine the same applies.