OpenCl: fix multi gpu support

Since #2268 the multi gpu support is broken. It is not possible to have more than one opencl context in a process.
Use one OpenCl context for the process instead per GPU.

update version to 2.9.0

cryptonight_r

merge back master -> dev

log level 4 as default

- add nvrtc to windows libraries
- add path to cuda.so for linux

- update coin definitions
- set rounding mode for cn_gpu
- remove currency cryptonight_wow (currently untested)

Co-authored-by: psychocrypt <psychocryptHPC@gmail.com>
Co-authored-by: fireice-uk <fireice-uk@users.noreply.github.com>

implementation is derived from the reverence implementation in xmrig

implementation is derived from the reverence implementation in xmrig

Co-authored-by: psychocrypt <psychocryptHPC@gmail.com>
Co-authored-by: fireice-uk <fireice-uk@users.noreply.github.com>

implementation is derived from the reverence implementation in xmrig

- set log lvl 4 as default
- increase the hashrate report time from 1min to 5min

release 2.8.3

set cn_gpu for Ryo

- remove fork for Ryo from cn_heavy to cn_gpu (now always cn_gpu)
- add example pool for ryo

increase version to 2.8.3

OpenCL: fix Vega invalid shares (windows)

CMakeLists.txt: Adjust OpenCL probing to find OCL SDK Light

Please fix Zelerius iterations

Implement cn-conceal

Vega GPUs create still invalid shares on windows 19.X and 18.7+ driver for cn_v8 and cn_heavy.

Add `OCL_ROOT` envvar support for Windows OCL SDK Light autodetection and prefer it over AMD APP SDK if both are installed

fix bfactor


fix

release 2.8.2

increase version to 2.8.2

OpenCL: fix the fix for the blockchain driver

 #2229 was not solving the issues

- revert #2229
- introduce the working fix

OpenCl: fix blockchain driver support

The OpenCl version of the blockchain driver is not understanding if
apointer to a pointer points into shared memory and throw an error
during the compilation.

- revert the usage of the struct to group all shared memory arrays

release 2.8.1

update algorithm documentation

- add more information if a algorithm derives from an other
- update `pools.tpl`

update version to 2.8.1

OpenCl: optimize cn_gpu