use position and data from chroma unit when treeType = DUAL_TYPE_CHROMA to derive context

Fix the huge loss in rate control with the macro WCG_EXT enabled by In-Loop Reshaping by saving the unadjusted lambda for RD cost before encoding current CTU.
This commit has removed the macro as recommended.

- create own files for HRD classes
- rename HRD to HRDParameters
- create HRD class that can hold all HRD related parameters as base for encoder/decoder specific implementations

diverges from current draft, thus using EMULATION_PREVENTION_FIX macro to mark changes

Fix ticket #267 on wrong SAD calculation in xGetSAD_NxN_SIMD(...) if rcDtParam.subShift is not equal to zero and input size is equal to 4x4.

patch-by: Tobias Hinz

Added:
* RDPCM_Y/Cb/Cr
* BDPCM
* TileIdx
* MTSFlag
* LFNSTIdx
* RegularMergeFlag
* SMVDFlag
* SbtIdx, SbtPos
* MIPFlag
* ISPMode
* JointCbCr
* ComAlphaCb/Cr (cross component prediction)
Removed:
* IBCFlag, IBC is signalled as part of PredMode

- includes special handling for dps ID equal to zero (JVET-N0869)
- writing of parameter set can be enabled by parameter EnableDecodingParameterSet

Xiang Li authored 5 years ago and Karsten Suehring committed 5 years ago

Fixed the encoder bug which leads to incorrect BVd signaling. The incorrect BVd triggers exception on valid BV checking.

implements:
- splitting tiles into bricks
- raster-scan tiles
- uniform and non-uniform splits
- all tiles/bricks in one slice
- each brick one slice
- PPS syntax

missing:
- rectangular tiles
- slice header syntax

further todo:
- rework slice mode 2 to use multiple tiles for byte size matching
- lots of renaming (tile->brick)

Training based on VTM 5.0rc1 CTC bitstreams

- The indexing of cost array was corrected by following shuffling pattern of full-RD candidate list employed by MIP.
- ~0.1% coding gain inc. along with ~4% encoding time dec. in AI

This reverts merge request !532

Deprecated function is removed as well as code calling it.
Impact on CTC: 0.00% (AI) and -0.01% (RA)

- The indexing of cost array was corrected by following shuffling pattern of full-RD candidate list employed by MIP.
- ~0.1% coding gain inc. along with ~4% encoding time dec. in AI