Isolating the important things from zero's obnoxiously large pull request

[ExpHP]

Labeling this as Help wanted. Take a look at #63 . This same branch mixes a lot of extremely important updates - like support for Touhou 19 and 20tr - alongside more controversial changes that I'd really need to discuss with him.

He added virtually no tests.

Here's some things that should not be too difficult for newcomers:

• Added acos/asin/tan. (Fixed in a couple of easy things pulled out from zero's PR #71)
• Fixed a critical bug in TH08 where the scratch registers are incorrectly typed. Just look for where "scratch" is mentioned. (Fixed in a couple of easy things pulled out from zero's PR #71)

Things I can probably mentor on:

• ending support

Some things that's less easy

• More flexible integer formats
  • IntFormat in the AST. (note: i want unsigned to be signed) ((WIP) Signed integers #72)
  • LLIR representation, formats for mapfiles
• th20 has a new field to std, he called it "unk", I'd rather __unknown1
• Added support for th18.5, th19, th20
• ids were added to std. This lets 8 STD files roundtrip that used to warn about "non-sequential id"s. Want to look over its behavior and make sure it has decent integration tests.
• ????????
• ????????
• ????????
• really there's a lot here, i need help pulling this apart

[ExpHP]

I checked off two of these boxes without merging the commits responsible for fixing them. Now I'm not sure where they are. WTF.

[ExpHP]

Okay all the things checked off now are actually merged

[ExpHP]

Update: I've been trying to make zero's PR easier to merge by rebasing it.

There is something very wrong with this branch.

Git is doing things I've never seen it do before. I do git rebase -i HEAD~20 and get conflicts. How does a branch conflict with its own history?!

I try to squash the first four commits of this branch, git rebase pauses before the squash and forces me to resolve conflicts without that commit included, and you can actually see a squash at the very top of the rebase TODO list, which I thought was impossible.

I tried to create a manual squash. I thought. I'll check out the merge base of the two branches at 1f77ccc, into a new commit. Then I'll replace the source code wholesale with the content of the fourth commit in the branch and commit that on top of it. Well, src/formats/ecl/ecl_06.rs exists in 1f77ccc and it does not exist in format_improvement_tests. I went even further back to 520368d, the commit that added the file. That commit is also in the history of zero's first commit. THE COMMIT THAT CREATED THE FILE IS IN THAT COMMIT'S HISTORY, HOW IS THE FILE NOT THERE, i want to cry

So, uh.............. I don't think I ever want to merge this even if I did approve it all. Because then I'll have a repository where git behaves like it's taken acid.