[yeqing-wu]

cost_prediction_mode() left out the adaptive MVD flag cost for same-ref compound modes and for TIP NEWMV, even though the flag is coded for every mode that allow_amvd_mode() accepts. The mode search tries both use_amvd values for these modes, so both were priced the same and the choice ignored the flag's rate.

Add the cost to those two returns, and only count it when enable_adaptive_mvd is on, which is when the flag is actually coded.

Anchor: commit 6169621
Speed 0 (cpu-used=0): VCWG CTC (130 frames, RA and LD)

1) RA
  +------------+-------+-------+-------+-------+------+------+
  | Class      |     Y |    Cb |    Cr |  wAvg | Enc% | Dec% |
  +------------+-------+-------+-------+-------+------+------+
  | A1         | -0.10 | -0.23 | -0.27 | -0.11 |  100 |  100 |
  | A2         | -0.14 |  0.03 | -0.19 | -0.13 |  100 |   99 |
  | A3         | -0.11 | -0.10 | -0.33 | -0.12 |  101 |   99 |
  | A4         |  0.04 |  0.02 |  0.47 |  0.05 |   99 |   99 |
  | A5         | -0.05 | -0.32 |  0.15 | -0.05 |  100 |   99 |
  | B1         | -0.06 | -0.15 |  0.02 | -0.06 |  100 |   99 |
  | B2         | -0.08 | -0.52 |  0.25 | -0.08 |  100 |  100 |
  | Avg w/o B2 | -0.09 | -0.09 | -0.09 | -0.09 |  100 |   99 |
  +------------+-------+-------+-------+-------+------+------+

2) LD
  +------------+-------+-------+-------+-------+------+------+
  | Class      |     Y |    Cb |    Cr |  wAvg | Enc% | Dec% |
  +------------+-------+-------+-------+-------+------+------+
  | A2         | -0.18 | -0.37 | -0.18 | -0.19 |  100 |  100 |
  | A3         | -0.28 | -0.52 | -0.28 | -0.29 |  100 |  100 |
  | A4         | -0.28 | -0.01 |  0.15 | -0.25 |  102 |  101 |
  | A5         | -0.50 | -0.08 | -0.96 | -0.50 |  101 |  101 |
  | B1         | -0.12 |  0.18 |  0.34 | -0.09 |  100 |  100 |
  | B2         |  0.05 |  0.02 |  0.52 |  0.07 |  101 |  100 |
  | Avg w/o B2 | -0.23 | -0.21 | -0.11 | -0.22 |  100 |  100 |
  +------------+-------+-------+-------+-------+------+------+

STATS_CHANGED

[yeqing-wu]

Hello all, this PR is ready for review, could you please help review the code?

[urvangjoshi]

@leolzhao : could you review this one, as it's related to AMVD?

[yeqing-wu]

@leolzhao and @yunqingwang1 , could you please help review this PR?