Optionally Pull CPython binary from InfluxData

[crepererum]

What

Provide a way so we can pull the CPython WASM binary from an InfluxData fork.

Why

@crepererum - I see you checked off the Bundle Python Guest #17 task (fine, you're doing it).

However, we're going to need to keep the embedded python up to date for CPython security (and bug) fixes and in looking at https://github.com/influxdata/datafusion-udf-wasm/blob/main/guests/python/Justfile#L62C7-L62C59, I see that we are pulling in compiled wasm from https://github.com/brettcannon/cpython-wasi-build, which is not very "official looking". I suggest at a minimum forking this repo to https://github.com/influxdata/cpython-wasi-build so we have a degree of control on the supply chain. Alternatively (and perhaps you've already done this), we could look for something more official. Can you add this as another subtask?

Originally posted by @jdstrand in #4

Note though that there's a bit of a conflict of interest here: on one hand InfluxData wants a forked version of the CPython WASM builds, on the other hand we want to eventually move this repository here to a datafusion-contrib (i.e. under a neutral umbrella).

How

Make the CPython fetching abstract enough so it can fetch from whatever source the user wants (given a URL and a checksum).

[jdstrand]

Make the CPython fetching abstract enough so it can fetch from whatever source the user wants (given a URL and a checksum).

This sounds reasonable for our purposes. I wonder if upstream datafusion wants their own fork and we could just use that?

[crepererum]

https://github.com/brettcannon/cpython-wasi-build is THE de facto standard binary for WASM at the moment. Quoting https://github.com/psf/webassembly

CPython from Brett Cannon (to experiment with what might eventually come from python.org)

So I don't see why an OSS org like Apache would need to maintain their own builds.