Initial Work on Performance Patches

[willtebbutt]

This PR starts to address #249 . In particular it

1. sets up conventions which ought to be followed for how these rules are to be incorporated into Mooncake, and
2. adds some initial rules.

This will make a couple of specific cases more performant, and make it possible for others to contribute rules that they would like to see in Mooncake.jl.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Files with missing lines	Coverage Δ
src/Mooncake.jl	100.00% <ø> (ø)
src/interpreter/abstract_interpretation.jl	87.50% <100.00%> (-1.24%)	⬇️
src/interpreter/ir_utils.jl	87.50% <100.00%> (ø)
src/rrules/performance_patches.jl	100.00% <100.00%> (ø)
src/test_utils.jl	93.07% <100.00%> (ø)

... and 1 file with indirect coverage changes

[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 │   Primal │ Mooncake │  Zygote │ ReverseDiff │  Enzyme │
│                     String │   String │   String │  String │      String │  String │
├────────────────────────────┼──────────┼──────────┼─────────┼─────────────┼─────────┤
│                   sum_1000 │ 100.0 ns │      1.8 │     1.1 │         5.5 │    8.21 │
│                  _sum_1000 │ 941.0 ns │     6.76 │  1420.0 │        33.5 │    1.07 │
│               sum_sin_1000 │  6.55 μs │     2.22 │    1.68 │        10.5 │    1.99 │
│              _sum_sin_1000 │  5.17 μs │     2.69 │   257.0 │        13.6 │    2.45 │
│                   kron_sum │ 267.0 μs │     49.5 │    12.2 │       263.0 │    10.3 │
│              kron_view_sum │ 614.0 μs │     22.7 │    5.12 │       125.0 │    49.3 │
│      naive_map_sin_cos_exp │  2.14 μs │     2.19 │ missing │        7.21 │    2.34 │
│            map_sin_cos_exp │  2.12 μs │      2.4 │    1.45 │        6.11 │    2.92 │
│      broadcast_sin_cos_exp │  2.25 μs │     2.26 │    2.27 │        1.49 │    2.27 │
│                 simple_mlp │ 413.0 μs │     4.84 │     3.1 │        6.71 │     3.5 │
│                     gp_lml │ 546.0 μs │     4.74 │     2.4 │     missing │    2.17 │
│ turing_broadcast_benchmark │  2.03 ms │     3.17 │ missing │        27.1 │ missing │
│         large_single_block │ 390.0 ns │     4.03 │  4440.0 │        31.3 │    2.18 │
└────────────────────────────┴──────────┴──────────┴─────────┴─────────────┴─────────┘

[willtebbutt]

Something unfortunate is happening with some allocation checking functionality. It's not really related to the main bit of this PR, but I'll need to fix it before merging.

[willtebbutt]

Latest commit purely contains changes to comments, so I'm going to merge as soon as the quality checks pass.