feat: check and warn about dirty git index during deploy

[zevisert]

See #56.

Not sure if the wording is alright with my new errors and logs -- please advise!

[zevisert]

This PR tries to minimize it's footprint. There's a bunch of other places (~5) that call get_app_version. Instead of shiming this new dirty check in there, I could also make a new function to verify the index is clean, and only call that from model.remote_deploy, thoughts?

Switched to explicitly opting out of get_app_version index checking at the callsites.

[zevisert]

After some reflection, I set the default value to True so that the API behaviour is the same (eg. all get_app_version call-sites need to opt-in to this check). Not sure what your preference is. The unionml cli is the only place that inverts this now, --ignore-diff is used to opt out of the check.

[cosmicBboy]

my inclination is the stricter the better... ignore_diff = False by default will blow up a deployment attempt unless the use specifically wants to allow uncommitted changes.

[zevisert]

I can do that, but #56 is only about failing the deployment. get_app_version is used in a bunch of other places, and in some of them it doesn't make sense to fail over a dirty index (eg: commands like list-model-versions). It's almost always the app_version = app_version or get_app_version() constructs that I think should opt out of this check.

Besides, in the current state of this PR, get_app_version gives an action hint on failure talking about --allow-uncommitted, which isn't a flag that is present everywhere that get_app_version is called. I'll update to allow_uncommitted = False, but to keep the rest of the functionality the same, I'll opt those other callsites out for now

[cosmicBboy]

awesome! changes look good, looks like pre-commit is having issues with convert-myst-to-ipynb... what version of jupytext do you have? Need to pin this in another PR to avoid this issue...

[zevisert]

what version of jupytext do you have?

> pyenv shell 3.7.13/envs/unionml
> jupytext --version 
1.14.0

[cosmicBboy]

re: pre-commit error, you'll need to rebase onto current main

[cosmicBboy]

hmm, something funny's going on with pre-commit, still investigating...

[cosmicBboy]

k, rebasing onto main should fix the issue

[cosmicBboy]

lgtm, thanks @zevisert !