switched default value for CodingStructure::initStructData::QP from -1 to MAX_INT (signalling that the QP value for the CS should not be set)

This reverts merge request !107

Fix ticket #116 on encoder crashes when AFFINE_MRG_MAX_NUM_CANDS is set to less than 5. Using constants AFFINE_MRG_MAX_NUM_CANDS and MRG_MAX_NUM_CANDS instead of fixed numbers in encoder checking.

Relates to !105

Fixes #107

Fixes #114

Fixes #108

When pruned == false and currCnt == MAX_NUM_HMVP_CANDS,
memmove() copied element motionCand[MAX_NUM_HMVP_CANDS] which is
out of bounds. With this fix one fewer elements is copied.
The element that was previously incorrectly copied was overwritten
by memcpy(). Hence there is no change in the result.
See also #113

Fixes #113

Fix ticket #110 on incorrect affine mv and mv predictor precision is used in GBI. Patch provided by Daniel.Luo@InterDigital.com.

This avoids running into issues such as the ones fixed in !88.
Formatting is also improved

Fix output of bit and tool statistics. If RExt__DECODER_DEBUG_BIT_STATISTICS and RExt__DECODER_DEBUG_TOOL_STATISTICS are enabled, the decoder throws and exception as size of enum CodingStatisticsType and the size of the statNames array don't match. Further, the order of the enum has to match the order of the labels. Also, the bit counter for JVET_L0283_MULTI_REF_LINE was missing.

Fix ticket #104 on incorrect position derivation In setMaxMinDepth. Fix provided by yj1003.piao@samsung.com.

Fix ticket #103 on MV rounding issue during AMVP candidates generation. Patch provided by ruoyang.yu@ericsson.com

add corresponding tool macros for CPR code, to prevent compiling errors when some tools are disabled

Fixes #102

Fix provided by zhi.a.zhang@ericsson.com
Fixes #100

Values were trained on CTC set (including optional tests) using
commit 67bafee8

Fix of ticket #98 on incorrect derivation of chroma-tree split context. Patch provided by adam.wieckowski@hhi.fraunhofer.de