Make MDC/O3 Replay GRB events more realistic

Currently we create mock GRB events as a part of internal MDC testing in first2years_external.py but only make background triggers. These are not completely realistic, not representing foreground triggers, by construction and could be improved in the following way:

• Make localization coordinates (RA, dec) more realistic by offsetting from injection coordinates. If pre-determined injection coordinates are not available, these could be approximated using the highest (density) probability pixel.
• Filter O3 replay events to only use BNS events. This could be done by checking P-astro (higher latency) or chirp mass as an approximation (lower latency).
• Potentially filter based off of inclination (viewing angle less than 25 deg?) if possible.
• Make sure if we have multiple GRB detections within the same time window, they are from different pipelines.

Note that both goals would either require knowledge that would not be available at the creation of a superevent (sky localization; p-astro) or would need a pre-determined injection set of the parameters (ra, dec, gpstime, masses/p-astro).