Set upper limit on findings from SAST analyzers to reduce memory consumption

After analysis in https://gitlab.com/gitlab-com/gl-infra/capacity-planning-trackers/gitlab-com/-/issues/2598#note_2948395369 we found that a single project generated a large SAST report that is over 250MB and contains over 250K findings.

This seems to be responsible for periodic memory spikes on the production "low-urgency-cpu-bound" shard and we would like to understand if these reports can be made to have an upper limit to reduce memory pressure on this fleet.