Removes the last vestigial parts of IBDI.  Currenltly, when FULL_NBIT=0,
g_uiBitDepth=8 and g_uiBitIncrement is luma_bit_depth_minus_8.

Replace with g_bitDepth that expresses luma_bit_depth.  Where bit_depth_minus_8
is required, derive from g_bitDepth.

This patch maintains the previous behaviour of FULL_NBIT while replacing
most of the FULL_NBIT guards.  When FULL_NBIT=1, distortion costs are
calculated using the full sample range.  When FULL_NBIT=0, distortion costs
are calculated as if using 8-bit source data.

From: David Flynn <dflynn@rim.com>