Define likely() and unlikely() macros to guide branch prediction.
Implementation requires GNU tool-chain (GCC, GNU ld).

Optional -D KROKI_LIKELY_REPORT[=RATIO] during compilation will enable
prediction analysis.  RATIO should be a float between 0.0 and 1.0,
default is 1.0.  Report on likely()/unlikely() with prediction rate
less than or equal to RATIO will be printed to stderr on program exit.
Prediction accounting is not thread-safe: when several threads try to
increment a counter simultaneously some increments may be lost
(however this should be a rare event and should not affect overall
statistics).

Defining KROKI_LIKELY_NOPOLLUTE will result in omitting
likely()/unlikely() alias definition (functionality will still be
available as kroki_likely()/kroki_unlikely()).

Inspired by Ulrich Drepper's "What Every Programmer Should Know About Memory"
(http://www.akkadia.org/drepper/cpumemory.pdf) annex A.2.