amo: apply fast-path analysis and optimization

[minsii]

Direct AMO mode still shows about 2x overhead than that of SOS

Benchmark: osu_oshm_atomics_all2one (shmem_int_finc -> MPI FOP)

        #SOS   #direct-amo
Theta/np=2 	14.88	39.22
Cori/np=2 	2.67    4.21

The current atomics check in OFI can be the cause. We need analyze the atomics path and apply optimizations similer to previous RMA (IPO inline, reduce instructions on AMO path)