Split Benchee integration tests

[Will-thom]

Summary

• split large Benchee integration coverage into focused test modules
• move shared integration assertions into Benchee.IntegrationHelpers
• replace the slow max_sample_size sampling assertions with configuration-level coverage and make the outlier test deterministic under slower runtimes

Fixes #480

Validation

• mix format --check-formatted test/benchee_test.exs test/benchee/save_load_test.exs test/benchee/hooks_test.exs test/benchee/measurement_integration_test.exs test/benchee/profiling_integration_test.exs test/benchee/function_overhead_and_outliers_test.exs test/benchee/evaluation_warning_test.exs test/benchee/escript_test.exs test/benchee/max_sample_size_test.exs test/support/integration_helpers.ex
• mix test test/benchee_test.exs test/benchee/save_load_test.exs test/benchee/hooks_test.exs test/benchee/measurement_integration_test.exs test/benchee/profiling_integration_test.exs test/benchee/function_overhead_and_outliers_test.exs test/benchee/evaluation_warning_test.exs test/benchee/escript_test.exs test/benchee/max_sample_size_test.exs --slowest-modules 10
• mix test --slowest-modules 10

The validation was run in an isolated Docker container with Elixir 1.17.3 / OTP 27.

[Will-thom]

Rebased this branch onto the current main and resolved the test split conflict by moving the newer max_sample_size default integration coverage into the dedicated max_sample_size test module.

Local validation ran in an ephemeral Elixir 1.17 Docker container:

mix test test/benchee/max_sample_size_test.exs test/benchee_test.exs

Result: 33 tests, 0 failures, 2 excluded.

[PragTob]

Heyo, thanks a lot! 💚

I was a bit busy, I'll take a look at this and likely adjust it a little.

[PragTob]

Merged as #487 - thank you very much!