Speed up a loop in findCrpr

[RRozak]

This is done by:

• inlining a call of clkPathPrev(Vertex *vertex, int arrival_index)
• removing redundant conditions
• returning the result of clkPathPrev(const PathVertex *path, PathVertex &tmp) only by reference

This loop is repeated over 500 million times in case of black_parrot design. The changes mentioned above give ~3.5% performance gain on grt stage. I compared the times of 10 runs of grt stage only:	Without my changes	With my changes
11:55.64	11:32.40
11:53.50	11:31.19
11:52.95	11:30.97
11:54.99	11:23.12
11:42.94	11:15.42
11:51.53	11:19.78
11:54.56	11:31.04
11:51.72	11:34.29
11:56.03	11:31.32
11:51.72	11:25.89

[maliberty]

Issues or PRs should be filed with https://github.com/parallaxsw/OpenSTA if still relevant. This is effectively a fork (though not strictly for historical reasons).