Stop asserting on bundler-stats output in tests

[JuanVqz]

Closes #45

Summary

• Tests were failing on CI for certain Ruby versions (3.0, 3.1 then 2.5 and so...) because bundler-stats resolves different gem versions and dependency counts depending on the Ruby version
• Since rails_stats just passes through bundler-stats data without processing it, asserting on specific gem versions was testing bundler-stats behavior rather than rails_stats logic
• Updated tests to verify structure and gem names from bundler-stats without asserting on version-specific values
• Code stats, totals, schema stats, and polymorphic stats assertions remain exact (they are stable across Ruby versions)
• Console output test now strips the bundler-stats gem table before comparing, keeping the full fixture for documentation purposes

Test plan

• bundle exec rake test passes locally
• CI passes across all Ruby versions (2.5 through 3.4)

[JuanVqz]

Ruby 3.3/3.4 resolves:
simplecov-console (0.9.2)
minitest-around (0.5.0)
terminal-table 3.0.2 → unicode-display_width (< 3)
(no unicode-emoji)
Total Gems: 17

Ruby 3.0/3.1 resolves:
simplecov-console (0.9.5)
minitest-around (0.6.0)
terminal-table → unicode-display_width (< 4)
→ unicode-emoji (~> 4.1)
Total Gems: 18

The reason is that bundle install picks different gem versions depending on the Ruby version. newer Ruby gets newer gem versions with different dependency trees. For example, terminal-table on older Ruby
pulls in unicode-emoji as an extra transitive dependency, which adds one more gem to the total count and changes the dependency numbers for simplecov-console.

That's why the bundler-stats portion of the output can't be pinned to a single expected value across the CI matrix; also, the tests didn't check it previously.

[jeremysmithco]

Hey, this looks good! I had one optional suggestion.

(There are some other underlying issues in this library that I noticed that I'd like to talk about later that I think we could address relatively easily, and would make testing easier.)