fix(test): Use trusted Relay in trace metric test

[loewenheim]

The size of a trace metric, which is relevant for billing, is supposed to be computed by the first trusted Relay in the chain and then not recomputed by subsequent Relays. This ensures that we only bill on data sent to us by the customer, not on normalizations we apply.

This PR parametrizes the test_trace_metric_extraction test with different scenarios for external Relays to check that this works correctly for both trusted and untrusted Relays.

[Dav1dde]

I think it makes sense to look at the test_ourlog_extraction_with_sentry_logs and port over all the different modes.

@pytest.mark.parametrize(
    "external_mode,expected_byte_size_1,expected_byte_size_2",
    [
        # The values here are billing relevant metrics, do not arbitrarily change it,
        # these values are supposed to be static and purely based on data received,
        # independent of any normalization.
        (None, 18, 359),
        # Same applies as above, a proxy Relay does not need to run normalization.
        ("proxy", 18, 359),
        # If an external Relay/Client makes modifications, sizes can change,
        # this is fuzzy due to slight changes in sizes due to added timestamps
        # and may need to be adjusted when changing normalization.
        ("managed", 128, 459),
    ],
)

Then again this is just a straight up improvement ✅

[loewenheim]

I think it makes sense to look at the test_ourlog_extraction_with_sentry_logs and port over all the different modes.

That's actually a good idea, I wasn't thinking that far.