Inline vp9_get_coef_context() (and remove vp9_ prefix).

Inline vp9_get_coef_context() (and remove vp9_ prefix).

Makes cost_coeffs() a lot faster:
4x4: 236 -> 181 cycles
8x8: 888 -> 588 cycles
16x16: 3550 -> 2483 cycles
32x32: 17392 -> 12010 cycles

Total encode time of first 50 frames of bus (speed 0) @ 1500kbps goes
from 2min51.6 to 2min43.9, i.e. 4.7% overall speedup.

Change-Id: I16b8d595946393c8dc661599550b3f37f5718896