Use g_bit_depth during input validation

The libvpx encoder requires the input frames passed to it by the
application for encoding have the same bit-depth as codec bit-depth. If
the input bit-depth is less than codec bit-depth, then the application
must upshift the frame before passing it. The application may
communicate the actual input bit-depth via g_input_bit_depth so that
quality metrics are computed with reference to actual input.

As the input is expected to have same precision as codec bit-depth,
this commit modifies the input validator to use codec bit-depth.

Also updated API documentation to reflect these changes.

Change-Id: I9d2fe17b61f141f0ed47cf78f2e1f717a695541c