Clarify search results

[xxxserxxx]

Implements ticket #76. There are two cosmetic changes, and two significant behavior changes. Watch the "song matches" column:

1. The columns are renamed "artist matches", "album matches", and "song matches". As discussed -- it removes any confusion about the search page being a browser.
2. The column titles now include a count of the number of search results. As this is updated in real-time, it also serves as a sort of progress bar
3. Results are now presented gonic-style: every entity in a group has a name or title that matches the query. This means that, for Navidrome, no longer are albums included who's artists match; however, those matching artists will appear in the "artist matches" column, so the results are included -- they just are no longer duplicated.
4. The manual paging has been removed. Searching is now performed concurrently. Pages of 20 results (in each column) are fetched, in sequence, until the server returns no results.

In particular on number 3: searches are interrupted by other searches, so that if the user starts a new search while a previous search is still pulling results, the previous search will be interrupted, the results cleared, and a new search started.

This feature in particular has made me realize we need a mock OpenSubsonic API server that we can configure behaviors for, for testing. On my subnet, even with a media library of over 17,000 songs, both Navidrome and Gonic return results too quickly for me to easily simulate behaviors of slower networks. Similar to why I never saw an issue with album art load performance, it's hard for me to type fast enough to trigger the search interruption.

[xxxserxxx]

I keep forgetting to sign these commits. Sorry.

[xxxserxxx]

Grrrr. This one is going to have conflicts. When you get to it, let me know and I'll rebase and clear the conflicts.

[spezifisch]

Grrrr. This one is going to have conflicts. When you get to it, let me know and I'll rebase and clear the conflicts.

I'm not ready yet, but your small video looks pretty impressive as far as our search capabilities go. Could be a nice showcase for the v1.0.0 release notes.

This feature in particular has made me realize we need a mock OpenSubsonic API server that we can configure behaviors for, for testing. On my subnet, even with a media library of over 17,000 songs,

Definitely either that or a CI test to benchmark latency against a Dockerized Navidrome and Gonic.

Also I think I didn't clarify that I'm in fact in the same LAN as my music server. The server is not high-end, but it's not that slow.

[xxxserxxx]

Also I think I didn't clarify that I'm in fact in the same LAN as my music server. The server is not high-end, but it's not that slow.

So, that's really weird. Is your library significantly larger than mine? When I run Navidrome, it's in a container on an old 4-core AMD, 8Gb RAM, that's also running a ton of other stuff: Home Assistant, gonic, a zwave integration server, mpd, and Jellyfin; it's my NAS. I'm seriously overloading the thing -- it's using 1GB of swap space, and has about 100MB of free memory at any given time. One of the cores is constantly pegged at 100% by Jellyfin, which seems to be constantly indexing stuff, or I have it misconfigured.

I'm curious as to why my Navidrome seems to respond more quickly. I honestly noticed no lag on the cover art, even after you mentioned it (I only run Navidrome when I'm testing).