Parallel compact blocks bandwidth reduction or improvements #27740

Opening a tracking issue post #25258

To summarize a few potential strategies:

1. It may be beneficial to reduce/eliminate concurrent block fetches from high-bandwidth outbound peers. That's at tension with the potential block relay speedup for non-listening peers, but at least non-listening peers aren't as vulnerable to sybil nodes.
2. Could reduce number of concurrent to 2, instead of 3, for non-listening only, or both modes, to reduce worst-case overhead/funneling to listening nodes
3. Could enable low-bandwidth compact outbound nodes instead by initiating a compact block fetch when a low-bandwidth outbound peer offers a header
4. Could add timers/schedules to give every announce attempt a "fair chance" before issuing another getblocktxn without unduly slowing down block relay.

A number of these would benefit greatly from a "blockrequests.cpp" style refactor to make this all more understandable.