Set some acceptable deltas for the tx cost difference calculation

[noonio]

trafficstars

Presently, it reports differences https://github.com/cardano-scaling/hydra/pull/1700#issuecomment-2410364453 - but almost all of these are because of variations in the generators, not functional changes.

If possible, remove this noise by only displaying differences over a certain size; or just run the comparisions with the same seed?