Fix coerce in Infer.subsCheck

[johnynek]

it is very concerning that changing these directions (widening instead of narrowing) doesn't require changing any tests.

I should probably try to find some example that fails in the previous direction... I guess it is actually very rare for subsCheck to not be just unifying the code into a single type rather than instantiating.

Or it could be that we just don't have any way of failing after coerce since we just assume the types are right after that so changing them can't change the tests... but it seems like we should be able to find some code that didn't type before that does now?

[codecov-commenter]

Codecov Report

Attention: 1 lines in your changes are missing coverage. Please review.

Comparison is base (6565b3a) 92.02% compared to head (369bdbe) 92.09%.
Report is 1 commits behind head on master.

❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1060      +/-   ##
==========================================
+ Coverage   92.02%   92.09%   +0.07%     
==========================================
  Files          93       93              
  Lines        9883     9882       -1     
  Branches     2311     2325      +14     
==========================================
+ Hits         9095     9101       +6     
+ Misses        788      781       -7     

Files	Coverage Δ
.../src/main/scala/org/bykn/bosatsu/rankn/Infer.scala	96.83% <91.66%> (+0.58%)	⬆️

... and 6 files with indirect coverage changes

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.