#2229 was not solving the issues

- revert #2229
- introduce the working fix

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

Add seperate kernel to prepare the scratchpad memory.

Combine the shared memory for a hash within one struct.
Reduce the shared memory footprint per hash by 64 byte.

- rename variable names like `b` and `bb` to something with a little bit
of meaning.

Optimize cn_gpu

based on the suggestion from @xmrig https://github.com/xmrig/xmrig-amd/commit/db4e169f3a78f273abf89ea8cf5bba7eccf1490b

cryptonight_turtle is only cryptonight_v8 with a different scratchpad,
iteration and mask value.
We are using now the new machanism to describe such derived POWs.

A POW is now defined by a function `f` and three degrees of freedom `f(iteration, scratchpad, mask)`.
`f` is the base algorithm like `cryptonight, cryptonight_gpu`
An easy to pars snytax to write the full POW definition down is: `cryptonight_gpu:0x0000c000:0x00200000:0x001fffc0`

This change make it very easy to integrate the new trend of variate the
number of iteations or the scratchpad size without modifying the full
code.

@xmrig provided the information that the driver 19.2.1 for vega also
create invalid results if pragma unroll is used for the groestl algo.

- use the user defined unroll
- auto suggestion:
  - only tune for cn_gpu if this is the main user
currency (after a fork)
  - set unroll to 1 for cn_gpu

OpenCl kernel using a larger work group size than configured by the
user to increase the occupancy. Depending on the algorithm and device
the size is limited.

This PR fixes that the user was able to select a invalid work group size.

If comp_mode is used the code will not compile.

- fix compile issue
- fix wrong conditions to handle `comp_mode`

Windows driver creates wrong code if unroll is used.

Fix message with the maximal allowed worksize if cryptonight_gpu is
used.

Use the algorithm names from `cryptonight.hpp` instead if number within the OpenCL kernel.

- fix broken trutle coin
- fix non cn_gpu algorithms

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

Avoid that we do multiple lookups to `std::map` to find the OpenCL
kernel binaries.

- add helper method `GetAllAlgorithms()` to get all active POW
algorithms
- select max scratchpad memory size based on the dev pool and user
algorithms

In the current implementation the POW algorithm in dev pool section of a
currency will not be taken into account during the binary creation.
This PR changes the behavior and allow to create binaries for more than two POW algorihms.

Since #2080 bittube2 is broken.

- reintroduce special AES function for bittube2

The default value for interleave was wrongly set to 50.

Remove the value and take the devault from the default constructor instead of side channeling it from the json parser.

NVIDIA is using clang as device compiler so the reciprocal optimizations was disabled with #2104.

- re-enable optimized reciprocal calculation

Add an option to brute force intensity settings and lock in at the intensity with the highest hashrate.

- update decumentation of the `interleave` option to mention the side effect with `auto-tune`
- disable `interleave` auto adjustment if `auto-tune` is enabled
- jconf: add `auto-tune` as optional option

- fix broken compile: change used `ULL` to `UL` because `UL` is defined as 64bit
- fix memory copy to shared memory via vload8 (somehow it create wrong access)

The auto config generates for AMD devices now by default two threads per GPU.

- remove the savety 128MiB memory now only from the max available GPU memory not from the avaialble memory for one alloc call
- extend the memory documentation in amd.txt

Due to a wrong implementation clamp was not working.

use for non clang (Rocm) OpenCL a optimized reciprocal calculation without lookup table.

Co-authored-by: SChernykh <sergey.v.chernykh@gmail.com>

Disable compatibility mode if intensity is a multiple of worksize. In that case enabled compaibility mode will only slow down the miner.

If two threads are using the same GPU device the start time of each hash round is optimized based on the average time needed to calculate a bunch of hashes.

This way to optimize the hash rate was first introduced by @SChernykh. This implementation based on the implementation in xmrig but differen in the details.

- introduce a new config option `interleave`
- implement thread interleaving

Use `mul24` to speedup the scratchpad index calculation.

Co-authored-by: SChernykh <sergey.v.chernykh@gmail.com>

Add new striding index where the memory is chunked by the size of the work group (worksize).

Co-authored-by: SChernykh <sergey.v.chernykh@gmail.com>

small optimization for non cryptonight_v8 algorithms

- optimize division

optimize cryptonight_heavy diff