for the record: heat based sorting, driven by training runs

[John Rose]

P.S. Also FTR, I sent a related set of ideas to Lilliput.  The point
there is that the dynamics of certain PARTS of object headers can be
predicted by training run profiles, and there may be possible uses for
such profiles.  Those are far more speculative than the ideas of
field, method, and BB ordering.

for the record: header bit dynamics could possibly change after Leyden training runs
https://mail.openjdk.org/pipermail/lilliput-dev/2024-August/001819.html