Compare ?T to T with ==/!= operators

[foobles]

Closes #1332.
Notes:

• Does not currently work for comparing vector to ?vector. I have never used them, so I don't know how to implement this. It works otherwise.

[andrewrk]

Nice work! This is a non-trivial improvement to the core of the stage1 compiler.

I have an idea for how this can perhaps be done in a simpler way. Maybe you already tried it and ran into a problem. But here's the idea:

• Leave ir_analyze_bin_op_cmp mostly the same, with only a check near the top for is_equality_cmp and detecting that we have a ?T compared with T situation. (or maybe split into two functions with only 1 that does the check)
• In this case, call ir_analyze_test_non_null to check if the ?T is null. If the result is comptime known, then you can return a comptime bool right there, otherwise:
• Use ir_build_cond_br_gen to do a runtime branch on whether the ?T is non-null. You can use ir_create_basic_block_gen to make the then, else, and end blocks and ir_set_cursor_at_end_gen to switch to each basic block.
• In the null case, we have a comptime known comparison result; branch to the end block.
• In the non-null case, call the ir_analyze_bin_op_cmp logic to do a comparison on T and T (no optionals here). branch to the end block.
• In the end block, use ir_build_phi_gen to make a phi instruction, that selects the respective values from the then and else blocks.

I think you asked me about how to do this on IRC previously and I told you to create a new Gen instruction to do this logic. Sorry about that, sometimes it's hard to give good advice when I'm working on something else and don't have my head fully wrapped around the code.

Can you give this way of doing it a try? I think it will result in more "DRY" code with less room for bugs to creep up in the future.

[andrewrk]

The test cases look good. Can you also add coverage for optional pointers (the motivating use case)?

[foobles]

@andrewrk Definitely, I didn't think that was possible. I'd be more than happy to use battle-tested combinators 👍

Also, is ?*T == *T just not tested as of now? Because it definitely already works without my change.

Additionally, should I just make a new PR? Since I basically have to revert everything here?

[foobles]

I think I will. See you in the next PR!