add >4-byte propagation to fastpath

_From [bruen...@google.com](https://code.google.com/u/109494838902877177630/) on March 21, 2014 23:11:56_

With VS2013 adding a lot of 16-byte xmm operations we're ending up with more and more slowpath trips. Here's media_unittests:

Per-size slow path executions:
1-byte: 2181
2-byte: 14772046
4-byte: 51967603
8-byte: 42013966
Other: 83001209

1M+ per-opcode slow path executions:
5 or: 2376833
55 mov: 1449990
56 mov: 2047464
126 punpcklbw: 8943621
127 punpcklwd: 2097681
128 punpckldq: 1076644
134 punpckhbw: 3355585
135 punpckhwd: 2098978
140 movd: 33287546
141 movq: 25386635
142 movdqu: 22130849
143 movdqa: 10292012
204 pextrw: 20246709
293 movups: 22949032
303 movhps: 11063894

_Original issue: http://code.google.com/p/drmemory/issues/detail?id=1483_