cc_cubic: sync to the new specification of RFC9438 for TCP CUBIC.

[cc]

Summary

highlighted changes

1. change the type of these CUBIC windows to uint32_t from uint64_t
2. sync to the updated formulas in RFC9438 and carefully handles these units (segments or bytes)
3. remove un-used function reno_cwnd()

change detail in cc_cubic.h

1. Updated cubic_k with the new formula in the RFC. It additionally uses cwnd_epoch for calculating K.
2. Also updated these theoretical formula functions in userland.

change detail in cc_cubic.c

1. cubic_cong_signal() now just flag the event and let the epoch start/re-start at the beginning of the current congestion avoidance phase in cubic_ack_received().
2. Increase the cwnd only by (target - cwnd) / cwnd with a restrictive target formula from the RFC. This shall further smooth the cwnd growth, as observed in the default stack.
3. Remove the obsolete hack for probing W_max in early congestion avoidance phase if used an old ssthresh from hostcache.
4. Also remove the obsolete hack that aligns the concave region after the first congestion event.

reference: https://reviews.freebsd.org/D41715

Test Plan

Tests coverage is in this wiki page: testD49540

I added additional tests in this second wiki page: second testD49540

And found a test case that shows +10% performance improvement of CUBIC in shallow buffered routing testbed with fat BDP (1Gbps x 80ms).
highlighted charts is this:
{F120278156}