Tagfetcher: cache fetched covers

[ronso0]

Closes #11084

A quick test to cache the fetched cover images with the global QPixmapCache a QMap.
Cache keys are "DlgTagftecher" + the MusicBrainz release id.
This greatly improves UX when comparing covers of MusicBrainz results.

Works good, though removing the covers from the cache is a bit dirty I think.
To discuss:
* really use global QPixmapCache which is also used by CoverArtCache?
* shall we increase the QPixmapCache cache limit when DlgTagFetcher is constructed to avoid clearing cached covers by adding large fetched covers?

I didn't check the sizes of fetched covers, yet, we need to estimate a reasonable limit.

ToDo

• clear cache when loading another track
• preferred image size vs. cached images:
  currently the preference is read from config each time image urls are received. With the current implementation this may result in a mismatch of current size preference / cached images.
  We could
  a) read from config once in the constructor to avoid the issue in the first place
  b) read the preference each time urls are received and use the size preference (int) for the cache key
• remove/disable debug output

based on #12300

[ronso0]

Rebased, I adjusted & simplified related code a bit, e.g. loading a cover to DlgCoverArtFullsize.
It's working fine now.

However, (ab)using QPixmapCache is not the way to go: it's capped to 20 MB, and fetching some 1200px covers would flush all covers cached by the library table.

So a QCache is unavoidable. (that simplifies clearing the flushing the fetched covers... a lot)

[ronso0]

Now the covers are cached in a QMap. I didn't set a cache limit, yet. Let's figure out a reasonable size, i.e. for how many high-res covers the size would become a problem.

I couldn't get it to work with QCache, I always got segfaults when retrieving pixmaps. Must be something with QCache taking ownership of the objects and/or with QPixmap being one of the implicitly shared classes, but I couldn't figure it out. Same when I stored the fetched QByteArray, it was always nulled for some reason. If someone wants to take a look, the experimental QCache commits are in https://github.com/ronso0/mixxx/tree/tagfetcher-cache-fetched-covers_qcache.

[daschuer]

Did you consider to extend use the exiting cover cache here?
I have not looked into it, but I can imagine that this way we can save one cover access.

An other option is to use QPixmapCache or the general purpose QCache container.

[ronso0]

In the comments above I explained why QPixmapCache is unsuitable (e.g. it requires manual cleanup), and that I didn't get it to work with QCache. QPixmapCache is derived from QCache).

Any idea how to fix the QCache implementation is welcome, even though the only benefits seem to be slightly faster lookup and automatic cleanup (limit).

[daschuer]

It looks like this is mergable after you have rebased out the debug commit.

[ronso0]

Oh great, will do.

[ronso0]

Done.
For TODO #2 I filed #13099

[daschuer]

Looks and works good! Thank you.