A bug in !1700 cause the encoder to pass all parameters into all layers (independent of the specified target layer)

At the encoder there are long parameters like "--QP=32" and short parameters like "-q 32"

For layer parameters -lx (with x being the layer index) the encoder always assumes that
-lx is followed by two parameters, which is only true for short type parameters.

This checks for "--" in the first parameter following -lx and only reads a second parameter,
if "--" is not found at the beginning.

Bugfix of JVETQ0172 (Disallow differing chroma format and different bit depths for cross-layer prediction) and bugfix of setting inter_layer_ref_pics_present_flag and constraint (When vps_independent_layer_flag[ GeneralLayerIdx[ nuh_layer_id ] ] is equal to 1, the value of inter_layer_ref_pics_present_flag shall be equal to 0.)

- catch std::bad_alloc explicitly
- remove catch-all handler

- remove HighPrecision MV option
- fixes for statistics output
- removal of SVN header integration (fixes issue with Xcode)