Fixes for issues affecting the FBref scraper

[lorenzodb1]

This PR fixes the following issues:

1. The filename for the cached team season stats was appended with "all" regardless of the type of stats queried. This caused an issue as the cache might not have contained the table needed. It now caches these tables in different files.
2. For every n rows, the website adds a row in a table that replicates the table header. This caused read_schedule to fail as the number of rows in df_table would be higher than the one of the list of match URLs obtained (see [FBref] Non-data rows in the table body should be removed #277). I added the logic to remove those replicated headers when found.
3. The website has no specific Scores & Fixtures on the Big 5 European Leagues Stats page. Thus it'd go to the generic Scores & Fixtures page, which shows games currently being played. Because of this, I had to move the optimisation that combines the top five leagues under that label in read_leagues, as read_schedule necessarily needs the five top leagues separately rather than in their combined form.
4. The method _fix_nation_col throws an IndexError, supposedly when no flag is present. I fixed this by changing the logic to use regular expressions instead so that when the flag is missing no error is thrown.

Additionally, it moves pretty-error to the dev dependencies group, as it would otherwise be installed in repositories importing this library (which should not be the case). I'm not sure I've done this correctly, and I had to remove some imports, so please let me know if this breaks previous behaviour and advise me on what I should do instead. It also updates pandas to v2.0.

[probberechts]

Thanks a lot. These are some great contributions. Could you just split them up into multiple pull requests? That makes it easier to review them + pull requests are used to automatically create the release notes.

[lorenzodb1]

Closing this PR as I'll break it down into individual PRs for each issue fixed.

#282
#283
#284
#285