Skip/Speed up tests, resolve most open issues

[mdavis-xyz]

I've had some issues open for a while. As discussed, it's hard to merge them confidently, given that the tests take half a day, and some functionality has broken due to changes to nemweb.

So I've worked to streamline the tests and the code itself. Now a test with an empty cache takes 2h30m. It's 1h40m with a full cache.

Some of the tests, I just skipped. I normally hate it when people do that. Although for now, I think it's worth doing temporarily, so that we can check all these other bug fixes and feature additions.

This fixes:

• Hard-coded cache path in test #65
• Set keep_csv=True as default for cache_compiler #63
• Excel file format cannot be determined #60
• Option to keep .zip not csv #56 (Now it keeps zips always)
• Tidy up incomplete files #55 (tidy up incomplete feather/parquer, but not yet CSV)
• Create cache location if it doesn't exist #51
• Add 1s data to DispatchPrice #44
• Is full data range necessary for tests? #16

And I include some currently open pull requests:

• Change registration list from xls to xlsx #61
• Use requests.Session for all .get downloads #48

I know it's bad style to combine so many unrelated issues into one pull request. However given that the master branch is not easily buildable, some of these do work together. Look through this pull request one commit at a time, if you want to see review each change separately.

Aside from the tests now skipped with @unittest.skip(), all unit tests pass.

I increased the minimum Python version to 3.10, from 3.9, because uv complained about a dependency. 3.9 is now deprecated, so I think this is ok.