Import matrix exponential rule from ChainRules

[willtebbutt]

I'm really not sure that this is an optimal implementation, but it does at least ensure that people can differentiate the matrix exponential.

[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/rrules/linear_algebra.jl	100.00% <100.00%> (ø)

[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 │     70.1 │     1.0 │         5.5 │ missing │
│                  _sum_1000 │     6.75 │  1460.0 │        33.0 │ missing │
│               sum_sin_1000 │     2.27 │    1.69 │        10.8 │ missing │
│              _sum_sin_1000 │     2.56 │   288.0 │        12.6 │ missing │
│                   kron_sum │     55.8 │    3.61 │       190.0 │ missing │
│              kron_view_sum │     59.5 │    9.41 │       244.0 │ missing │
│      naive_map_sin_cos_exp │     2.53 │ missing │        7.12 │ missing │
│            map_sin_cos_exp │     2.72 │    1.43 │        6.04 │ missing │
│      broadcast_sin_cos_exp │     2.62 │    3.14 │        1.45 │ missing │
│                 simple_mlp │     6.72 │    3.16 │        10.1 │ missing │
│                     gp_lml │     9.08 │    3.98 │     missing │ missing │
│ turing_broadcast_benchmark │     3.17 │ missing │        28.2 │ missing │
│         large_single_block │     4.03 │  4530.0 │        30.4 │ missing │
└────────────────────────────┴──────────┴─────────┴─────────────┴─────────┘