Run CI on julia nightly only

[IanButterworth]

Testing along side #142 to see if codecov reports are the same as master when only julia nightly is tested

[codecov]

Codecov Report

Merging #143 (d1e97a1) into master (db713f3) will decrease coverage by 12.63%.
The diff coverage is n/a.

@@             Coverage Diff             @@
##           master     #143       +/-   ##
===========================================
- Coverage   88.72%   76.09%   -12.64%     
===========================================
  Files          13       13               
  Lines         878      937       +59     
===========================================
- Hits          779      713       -66     
- Misses         99      224      +125     

Impacted Files	Coverage Δ
src/global_constants.jl	18.00% <0.00%> (-51.39%)	⬇️
src/funcptrs.jl	55.40% <0.00%> (-44.60%)	⬇️
src/matmul.jl	78.48% <0.00%> (-14.35%)	⬇️
src/block_sizes.jl	74.11% <0.00%> (-12.79%)	⬇️
src/memory_buffer.jl	68.00% <0.00%> (-5.92%)	⬇️
src/init.jl	75.00% <0.00%> (-3.58%)	⬇️
src/complex_matmul.jl	100.00% <0.00%> (ø)
src/forward_diff.jl	98.11% <0.00%> (+0.33%)	⬆️
src/utils.jl	68.42% <0.00%> (+0.36%)	⬆️
src/macrokernels.jl	90.36% <0.00%> (+1.17%)	⬆️

---

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 db713f3...d1e97a1. Read the comment docs.

[chriselrod]

JuliaLang/julia#44625 (comment)
Do you know when/how often the nightly builds are rebuilt?

[IanButterworth]

After every commit to master, but only after the tests complete and pass, plus some delay for AWS hosting to update.. it's a little hard to predict, and there's no way to query the nightly commit beyond running CI like this or downloading and installing them locally.

Probably easiest to just retrigger this in the morning

[IanButterworth]

@chriselrod on a quick look it seems like the coverage diff might be real now? as in different code paths being hit on nightly?

[IanButterworth]

on a quick look it seems like the coverage diff might be real now? as in different code paths being hit on nightly?

If that holds true, my motivation here is satisfied. i.e. coverage tracking is functional and a lot quicker.

Who's a good person to decide how to tidy up the CI here to avoid duplication? I don't really know Octavian