Handle payload superclass redefinition errors during gem RBI validation

[zeeshankhan-05]

Motivation

Fixes #1834.

When Tapioca validates generated gem RBIs, some definitions can redefine a Sorbet payload class with a different superclass. Sorbet reports this as error 5012 and recommends adding --suppress-payload-superclass-redefinition-for=ConstantName to sorbet/config. Tapioca already auto-adjusts RBI strictness for error 4010, but it did not handle 5012 or inform users about the mismatch.

Closes #1834

Implementation

• Keep the existing --stop-after namer validation pass for error 4010 auto-strictness behavior.
• When auto_strictness is enabled, run a second full Sorbet pass to detect resolver-level error 5012 (the namer pass alone does not surface this error).
• Parse the affected constant from Sorbet error output and append --suppress-payload-superclass-redefinition-for=ConstantName to sorbet/config when missing.
• Use exact-line idempotency (config.lines(chomp: true).include?(flag)) so similar suppressions for other constants do not block the correct entry.
• Preserve existing sorbet/config content when appending (no stripping of trailing whitespace or blank lines).
• Print a user-facing message when a suppression is added or when it is already present, so the mismatch is not fixed silently.

Tests

Added an integration test in the describe "strictness" block of spec/tapioca/cli/gem_spec.rb (following maintainer guidance on #1834). The test exercises tapioca gem, verifies the suppression is written to sorbet/config, verifies user-facing output on first and second runs, and confirms idempotency.

Ran locally on the clean branch (1 commit ahead of upstream/main):

• bin/test spec/tapioca/cli/gem_spec.rb — 67 tests, 494 assertions, 0 failures, 0 errors, 1 skip
• bin/test spec/tapioca/cli/dsl_spec.rb — 69 tests, 335 assertions, 0 failures, 0 errors, 0 skips
• bin/typecheck — pass
• bin/style — pass

An earlier full-suite run (bin/test) completed with 794 tests, 0 failures, 2 errors, and 2 skips. The two errors were unrelated RubyLsp::Tapioca::Addon URI::InvalidURIError failures caused by spaces in the local directory path; they are not related to this change.

[zeeshankhan-05]

Hi @KaanOzkan. I appreciate the help with #1834 about using the describe "strictness" block in spec/tapioca/cli/gem_spec.rb and verifying suppression, user messaging, and idempotency. I followed that approach in this PR. Whenever you have some time, would you be able to review my PR?

[zeeshankhan-05]

I have signed the CLA!