refactor: update `tx speed` network stats to use median latency for the bucket with median load

[mistakia]

Use the median latency from the bucket that has a median number of blocks, instead of the overall median latency. This way, if one bucket has too many blocks, it won't skew the latency measurement for the rest. This method gives a more accurate latency expectation for most buckets.

[mistakia]

updated to use median latency of median bucket (by # of confirmed blocks) only for 24h period