Add more tests for textures

[Timic3]

In #11 we have gotten rid of dxt-js dependency and implemented our own decompression code. Since we handle decompression now, we need proper tests for this.

I suggest running a few textures (for each compression algorithm) through dxt-js (we don't need to run it through on each test, just use result as static data) and then compare. This way, if we find issues even with current implementation, we can fix them.

[AnriTool]

I started checking and preparing the test, and saw that sometimes the data may differ by 1 unit.
For example, RGB in dxt-js outputs 34, 180, 130, while the ts implementation outputs 34, 179, 130.

This may be due to rounding, can we consider a deviation of 1 unit as an error?

[Timic3]

I think 1 unit is fine, since it's not really noticeable. It might as well be that dxt-js might be wrong, since txd.py implements GTA's reversed code, if I'm not wrong?

Anyways, as far as I've tested on the rw-viewer (which I will open source soon), the textures on the buildings, cars and skins look fine.