Macro compilation tests.

[Anders429]

This PR defines macro compilation tests using the trybuild crate. These tests are for every macro defined within this crate. Additionally, some macros (specifically, the stages! macro and a small amount of the entities! macro) were rewritten to match the expected output (which is good, that means the tests are working!). More tests could be added in the future with #54, but for now I believe this covers all the possible failure cases of the macros.

[codecov-commenter]

Codecov Report

Merging #57 (1e8dcfa) into master (bec99bb) will decrease coverage by 0.73%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master      #57      +/-   ##
==========================================
- Coverage   22.00%   21.27%   -0.74%     
==========================================
  Files          54       54              
  Lines        3853     3986     +133     
==========================================
  Hits          848      848              
- Misses       3005     3138     +133     

Impacted Files	Coverage Δ
src/entities/mod.rs	0.00% <ø> (ø)
src/lib.rs	100.00% <ø> (ø)
src/world/mod.rs	0.00% <0.00%> (ø)
src/system/mod.rs	0.00% <0.00%> (ø)
src/system/par.rs	0.00% <0.00%> (ø)
src/query/result/iter.rs	0.00% <0.00%> (ø)
src/system/schedule/mod.rs	0.00% <0.00%> (ø)
src/query/result/par_iter.rs	0.00% <0.00%> (ø)
src/system/schedule/builder.rs	0.00% <0.00%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update bec99bb...1e8dcfa. Read the comment docs.

[Anders429]

There are two failures, caused by things outside of this PR:

• miri: seems that miri isn't set up to allow directory-level access, which is required by trybuild. Essentially, readdir64 is not yet included in the list of foreign functions supported. This appears to be the tracking function.
• valgrind: this appears to be a bug within cargo-valgrind. The output from valgrind doesn't match what cargo-valgrind is expecting. Doesn't seem to be caused by the tests at all.

[Anders429]

Fixed the cargo-valgrind issue by upgriding the valgrind version in the CI. Unfortunately, this leads to a new problem, which is that valgrind finds a memory leak specifically when running the trybuild tests. As far as I can tell, it's detecting a link directly with the trybuild TestCases object. This is not useful for the crate, and can be safely ignored.

Both of these failures seem to be out of scope of this crate. I'm thinking the best option moving forward is to just disable trybuild tests on the miri and valgrind jobs.

[Anders429]

Just to clarify, the decision to ignore these tests for valgrind and miri is fine because the only aspect that would be checked in either of those workflows is the trybuild::TestCases object. The compilation tests wouldn't be tested whatsoever, so we aren't even checking anything by running either tool over these tests.

I will be keeping the more recent version of valgrind, as I think that has benefit regardless of whether it is needed to make the workflow work correctly. Using up-to-date software is a good idea, and the work is now already done to make it up-to-date.