Reruns bundle exec srb rbi hidden-definitions, resolves typecheck

[mattxwang]

This PR fixes the problem on master, which was introduced with a new version of a gem (bundler? or maybe the runtime RBI reflection is not consistent, since the sorbet-runtime is the update I have when building from scratch) - our RBIs were out of date in-between when the CI check last ran and when we merged the PR.

I could see this problem happening more in the future: between the last CI check for a PR and it being merged, a gem is updated; bundler will resolve them differently (using the newer version on merge), and this could cause a mismatched RBI error.

Nominally, this fixes the problem, but this seems like a brittle issue (since we don't lock the gems we test against with a Gemfile.lock). Any thoughts on how to prevent this from being an issue in the future?

[codecov]

Codecov Report

Merging #66 (5870f88) into master (c3a374d) will increase coverage by 0.07%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master      #66      +/-   ##
==========================================
+ Coverage   99.14%   99.22%   +0.07%     
==========================================
  Files           6        6              
  Lines         350      386      +36     
==========================================
+ Hits          347      383      +36     
  Misses          3        3              

Impacted Files	Coverage Δ
lib/sorbet-coerce/converter.rb	97.36% <100.00%> (+0.22%)	⬆️
spec/coerce_spec.rb	100.00% <100.00%> (ø)