[P2P] Increase performance of Good

[Liquid369]

This PR actually opens up our need for some more fuzzing and testing to integrate while catching up to various Bitcoin improvements.

It is a partial back port of bitcoin#22974 , reasons being we have to isolate more of the fuzz integrations so we can add them into the code base. Although this provides an improvement of over a factor of 100 for the population of the addrman 'good' peers because we iterate the table twice in order to accomplish this.

First loop is modified because in MakeTried() we can speed this up by stopping it mainly at nRefCount == 0, any further it would absolutely hit an assert. And then we can also check from the bucket that it is originating from peer wise, giving us less buckets to check.
Second loop is removed because in Good_() we look for the entry but if it is not corrupted, there's two places it would be, in new, tried or not in addr db at all. Which means at this point we would not need this and can return earlier.

Some benchmarks from upstream bitcoin#22974 (comment) bitcoin#22974 (comment)

[panleone]

utACK 31e776d:
the variable nUBucket is no longer used so removing that for loop is a nice speedup,
the changes in MakeTried speed things up but the logic is the same

[Fuzzbawls]

ACK 31e776d