Range requests

[danielballan]

As noted in #734 the motivating use cases here was DuckDB's httpfs, which tries to do partial access of files over HTTP. Given a server with one DataFrame in it:

tiled serve catalog --temp

from tiled.cilent import from_uri
c = from_uri('http://localhost:8000', api_key='secret')
df = pandas.DataFrame({"A": [1,2,3], "B": [4,5,6], "C": [7,8,9]})
c.write_dataframe(df, key="x")

DuckDB's httpfs plugin issues requests that use the Range header:

import duckdb
db = duckdb.connect(':memory:')
db.execute('install httpfs; load httpfs;')
db.execute('SELECT * FROM read_parquet("http://localhost:8000/api/v1/asset/bytes/x?id=1&api_key=secret")').fetchall()

generating these server logs confirming that Partial Access was invoked:

[531d1958246ae636] 127.0.0.1:59854 (public) - "HEAD /api/v1/asset/bytes/x?id=1&api_key=secret HTTP/1.1" 405 Method Not Allowed
[1e2951b2eaa982dd] 127.0.0.1:59854 (admin) - "GET /api/v1/asset/bytes/x?id=1&api_key=secret HTTP/1.1" 206 Partial Content
[6a5b9f06acf0a7e5] 127.0.0.1:59854 (admin) - "GET /api/v1/asset/bytes/x?id=1&api_key=secret HTTP/1.1" 206 Partial Content
[79d13fc32a84235e] 127.0.0.1:59854 (admin) - "GET /api/v1/asset/bytes/x?id=1&api_key=secret HTTP/1.1" 206 Partial Content
[90ebff815c556fa1] 127.0.0.1:59866 (admin) - "GET /api/v1/asset/bytes/x?id=1&api_key=secret HTTP/1.1" 206 Partial Content

Checklist

• Add a Changelog entry
• Add the ticket number which this PR closes to the comment section

Closes #734

[hyperrealist]

Looks good to me, with the caveat that I could only do a shallow review of file_response_with_range.py given my inexperience with starlette. But the logic checks out.

I edited file_response_with_range.py to replace literal http status codes with HTTP_200_OK et. al. from starlette.status.

Also there are few minor comments in the files.