Adds pyproject - #2117

I cannot figure out why the conda and docker builds are failing but otherwise this should be sufficient

I cannot figure out why the conda and docker builds are failing but otherwise this should be sufficient

It looks like you're missing pywavlets (pywt) and h5py in your test environment.

It's also skipping a lot of tests as the optional (but required for out CI) packages aren't there:

----------------------------------------------------------------------
TEST SYSTEM CONFIGURATION
CIL version:  24.3.1.dev25+g5e59972b
{'has_astra': False,
 'has_ccpi_regularisation': True,
 'has_cvxpy': False,
 'has_ipp': True,
 'has_matplotlib': True,
 'has_numba': False,
 'has_nvidia': False,
 'has_tigre': True,
 'has_tomophantom': True}
----------------------------------------------------------------------

I cannot figure out why the conda and docker builds are failing but otherwise this should be sufficient

It looks like you're missing pywavlets (pywt) and h5py in your test environment.

It's also skipping a lot of tests as the optional (but required for out CI) packages aren't there:

----------------------------------------------------------------------
TEST SYSTEM CONFIGURATION
CIL version:  24.3.1.dev25+g5e59972b
{'has_astra': False,
 'has_ccpi_regularisation': True,
 'has_cvxpy': False,
 'has_ipp': True,
 'has_matplotlib': True,
 'has_numba': False,
 'has_nvidia': False,
 'has_tigre': True,
 'has_tomophantom': True}
----------------------------------------------------------------------

Ok, I saw the missing packages pywt and h5py but didn't notice this, so I thought it was just those 2 packages which confused me(the though it's still a bit confusing as matplotlib still seems to be installed).
All of the packages should be installed and I'm not quite sure why they aren't. I'll look into this again later.

Ok, so I think I see the problem now.
Since the purpose of the conda CI is to work in an isolated conda environment, I can't just install stuff into the environment externally(both technically without hacks and conceptually).
conda doesn't seem to have the capability to work with pyproject.toml files(see conda/conda#12462), and, if I understand conda correctly, this dependency information also needs to be included in the recipe file.

A few options:

1. duplicate dependencies in recipe files and pyproject. I don't like this solution, but it is a short term solution, and could potentially be done as an intermediate step. It'll be very annoying to deal with while it's an intermediate step, since dependencies will have to manually be updated in multiple places.
2. Prepare the project so that it's publishable to pypi, publish to pypi, and then upload to conda-forge(as demonstrated here). To me this seems more logical, but does require a bit more work.
3. Use an experimental feature in rattler-build to read the pyproject file.

I'm not sure how to do this super incrementally; the work that's been done so far is reasonably close to making 2. possible, but I would definitely prefer a more incremental strategy if anyone has suggestions.

Nevermind: I think I found a way to get this to work with the tool grayskull. It can generate recipes from sdists, so I'll try that.