Support server content file manifest via manifest.txt to reduce number of guessing

[Crystalwarrior]

Right now, WebAO has to make ridiculous guesses as to what kind of file extension the char.ini is referring to, because AO is AO and the content is never referenced with its extension present.

For KFO, I created a batch tool to check if you're missing any files by comparing against a manifest.txt that I generate via commandline. The manifest.txt is generated like this:
dir/s/b > manifest.txt
Afterwards, I simply open it in a text editor (e.g. VisualStudio Code), Ctrl+F the absolute filepath, and replace it with empty space to make it all relative filepaths.

This same manifest can be used to fully understand which exact file path the char.ini is referring to, making sure that WebAO always only makes one call to the exact file extension that it knows in advance. This will also massively reduce latency, as it no longer has to guess a .gif, .webp or .apng, and it also will understand if Case Cafe-style (a)/ (b)/ folders exist (#84).

Server owners would have the responsibility of making sure the manifest.txt exists in the asset repository they have. Otherwise, the old guessing algorithm will be used.

[Crystalwarrior]

Here's the manifest txt I generated in the .gif
manifested.txt

(rename it to manifest.txt, I only named it manifested to not replace the manifest.txt I already have)

[oldmud0]

Hmm. This wouldn't be a bad idea after all.

[notcancername]

While talking about the possibility of supporting new formats, this came up:

(me): Perhaps WebAO should include an index? Just the output of `find` should already be a massive performance benefit.
[...]
(me): For the official client, the output of `find . -type f` is only 108K (2759 files).  This would be much more for a client like KFO, but even there the benefit from the index should outweigh the delay of making 3 HTTP requests for every animation.

I very much agree with this change. Making one request for every possible file extension to check for its existence is a massive hack caused by HTTP not having a proper mechanism for directory listings.

It worries me that this issue is apparently well-known (Salanto brought it up in response to an example addition of a format), but the obvious solution isn't implemented. I'm going to try and implement it.

[notcancername]

This is definitely not a good idea as-is. It's too big and will destroy users' mobile data plans. I'm looking into a better solution.