Use trace instead of protocol for lantern

[connorjclark]

See: #16026

ctrl+f INTERNAL_LANTERN_USE_TRACE

There are ~7 tests that produce different results when using the trace instead of the CDP network log. Each one needs to be investigated fully and understood. For the cases where the trace implementation has an error, we need to fix it (Lantern lives in CDT repo now).

debugNormalizeRequests is critical for debugging the differences.

It's probably not necessary to treat this as a breaking change, but it seems like the safe approach.