Run everything on 1.10

[willtebbutt]

#384 points out that the workaround I introduced in #377 only works on 1.11. I didn't catch this because, so far, I've only been running a subset of CI on 1.11 in order to keep CI times down. This PR will

1. change our CI set up so that all tests are run on LTS and release (preventing this kind of problem from recurring), and
2. fix the problem on 1.10.

If it looks like the amount of time spent running CI is becoming prohibitive again, I'll have to look at ways to speed it up.

[codecov]

Codecov Report

Attention: Patch coverage is 93.54839% with 2 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
src/interpreter/ir_utils.jl	0.00%	2 Missing ⚠️

Files with missing lines	Coverage Δ
ext/MooncakeLuxLibSLEEFPiratesExtension.jl	100.00% <ø> (ø)
src/interpreter/abstract_interpretation.jl	83.05% <ø> (ø)
src/interpreter/contexts.jl	100.00% <100.00%> (ø)
src/interpreter/ir_normalisation.jl	89.39% <ø> (+3.78%)	⬆️
src/interpreter/s2s_reverse_mode_ad.jl	92.29% <100.00%> (+0.21%)	⬆️
src/rrules/builtins.jl	97.93% <100.00%> (+0.71%)	⬆️
src/rrules/fastmath.jl	100.00% <ø> (ø)
src/rrules/misc.jl	97.36% <ø> (ø)
src/test_utils.jl	93.20% <ø> (ø)
src/interpreter/ir_utils.jl	87.50% <0.00%> (-1.67%)	⬇️

... and 8 files with indirect coverage changes

---

🚨 Try these New Features:

• Flaky Tests Detection - Detect and resolve failed and flaky tests

[github-actions]

Performance Ratio:
Ratio of time to compute gradient and time to compute function.
Warning: results are very approximate! See here for more context.

┌────────────────────────────┬──────────┬─────────┬─────────────┬─────────┐
│                      Label │ Mooncake │  Zygote │ ReverseDiff │  Enzyme │
│                     String │   String │  String │      String │  String │
├────────────────────────────┼──────────┼─────────┼─────────────┼─────────┤
│                   sum_1000 │     70.8 │     1.0 │        5.61 │ missing │
│                  _sum_1000 │     6.71 │  1190.0 │        35.6 │ missing │
│               sum_sin_1000 │      2.3 │    1.69 │        12.6 │ missing │
│              _sum_sin_1000 │     2.71 │   237.0 │        13.3 │ missing │
│                   kron_sum │     95.2 │    4.56 │       307.0 │ missing │
│              kron_view_sum │     58.9 │    9.38 │       218.0 │ missing │
│      naive_map_sin_cos_exp │     2.56 │ missing │        7.19 │ missing │
│            map_sin_cos_exp │     2.78 │    1.53 │        6.06 │ missing │
│      broadcast_sin_cos_exp │     2.59 │    2.25 │        1.47 │ missing │
│                 simple_mlp │     6.36 │    2.48 │        9.15 │ missing │
│                     gp_lml │     15.6 │    4.42 │     missing │ missing │
│ turing_broadcast_benchmark │     3.16 │ missing │        23.9 │ missing │
│         large_single_block │     4.01 │  4050.0 │        30.6 │ missing │
└────────────────────────────┴──────────┴─────────┴─────────────┴─────────┘

[willtebbutt]

This is getting there, but there are a few issues on 1.10 that need dealing with. I've made some progress today, and hope to finish off tomorrow.

[willtebbutt]

@lcontento it looks like we're nearly there -- down to one error. I have to call it a day now -- should be able to resolve in the morning!