feat(detectors): Update detection algorithm for MN+1 Experimental Detector

[roggenkemper]

this pr makes a few changes to the MN+1 Experimental Detector:

1. it updates the _equivalent method to differentiate between spans that have the "default" op , which is common in Prisma. This made some detected issues be slightly off because it was finding a "pattern" with two spans that weren't necessarily the same
2. It updates the definition of a valid pattern to exclude those that start with a serialization span. serialization spans come after a query or operation, so it doesn't make sense to start on them
3. it keeps more of the most recent spans when leaving the continuing state which should allow for more accurate pattern finding.
   these changes should give more accurate results and fix issues we saw where the pattern was starting/ending 1-2 spans from where we expected it.

[cvxluo]

nice test