Correctly handle PartialTypeVars in inference

[willtebbutt]

Should resolve the new problem in #282 .

Occasionally, the inference result associated to an SSA is Core.Compiler.PartialTypeVar. This means that inference has successfully determined that the SSA will be a TypeVar. More than that, it has figured out some of the runtime information about the SSA, in a similar sense (I believe) to the way that Core.Compiler.PartialStruct knows some runtime information about a given struct. I've asked on slack for a MWE of a function in which this is the result of inference. I'll wait a few hours, but will merge this with the current unit test if I nothing appears.

edit: many of the people who would know about this are based in the US. I'll get this merged asap, and make a new PR if an MWE becomes available.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Files with missing lines	Coverage Δ
src/interpreter/abstract_interpretation.jl	83.05% <100.00%> (+0.29%)	⬆️

[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 │     69.2 │     1.0 │         5.5 │ missing │
│                  _sum_1000 │     6.78 │  1190.0 │        37.0 │ missing │
│               sum_sin_1000 │     2.31 │    1.68 │        10.6 │ missing │
│              _sum_sin_1000 │     2.68 │   373.0 │        13.3 │ missing │
│                   kron_sum │     88.5 │    6.76 │       292.0 │ missing │
│              kron_view_sum │     58.9 │    9.48 │       242.0 │ missing │
│      naive_map_sin_cos_exp │     2.54 │ missing │        7.02 │ missing │
│            map_sin_cos_exp │     2.78 │    1.42 │        6.14 │ missing │
│      broadcast_sin_cos_exp │     2.57 │    2.25 │        1.45 │ missing │
│                 simple_mlp │     4.68 │    3.03 │        6.28 │ missing │
│                     gp_lml │     4.57 │    1.59 │     missing │ missing │
│ turing_broadcast_benchmark │     3.78 │ missing │        23.3 │ missing │
│         large_single_block │     4.13 │  4400.0 │        31.3 │ missing │
└────────────────────────────┴──────────┴─────────┴─────────────┴─────────┘