small optimization for non cryptonight_v8 algorithms

Reduce local memory foot print to increase the occupancy.

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

define shared memory in the outer scope

- optimize kernel cn0 and cn2
- optimize vast int math
- use more 32bit variables

Co-authored-by: psychocrypt <psychocryptHPC@gmail.com>

optimize the devision in cryptonight_heavy and cryptonight_haven

import of https://github.com/xmrig/xmrig-amd/pull/185/commits/5d9b9334654df25cea7707f667990fd1577ed290

In the current implementation the bit align is using signed integer which results in pulling in
ones in the case the sign bit is set.

- cast to unsigned integer before using bitshift

With rocm we fighted very long with invalid shares. This is now solved with rocm 1.9 and
this tiny fix.
It is not fully clear where a memory optimization is kicking in and break the kernel `Groestl` if the variables `M` and `H` are not `volatile`.
The performance ill not change with this fix.

The fix is tested with rocm 1.9 with a VEGA64 and a RX570

add cpu implementation for the final monero POW

- fix code style issues
- fix spelling issue
- fix asm to support newer clang versions

add option `unroll` for OpenCL to allow better tuning the main POW kernel.

Create a special pass for NVIDIA GPUs to load memory chunks first into the shared memory.

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

- implement cryptonight_v8
- update auto adjust to fit the special requirements of `cryptonight_v8`
- add fast math integer implementation for `sqrt`, `reciprocal`  and `division`

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

If the first bit of the nonce is `1` (this is very often if we use a nicehash pool)
than it could be that some OpenCL implementations handle the 64bit representation of the 32bit
nonce on the device side as signed integer.
During a right bitshift we pull wrong ones from the wrong higher part of the 64bit
nonce representation into the 32bit part of the nonce.
The result will be that the computed share is invalid.

- explicit cast the nonce on the device to `uint` to avoid any side effects

- add cryptonight_heavy derivate cryptonight_bittube2
- add coin bittube
- remove coin ipbc because this coin is now called bittube

- explicit loop unrolling

based on changes in @imperdin fork https://github.com/imperdin/xmr-stak/blob/master/xmrstak/backend/amd/amd_gpu/opencl/cryptonight.cl

- update pools.txt
- add new algorithm `cryptonight_haven`
- update all backends

- rename cryptonight_fast to cryptonight_masari
- set dev pool to cryptonight_monero

solve #1494

- add algorithm `cryptonight_v7_stellite` (internal named: `cryptonight_stellite`)

- add algorithm `cryptonight_lite_v7_xor`
- update documentation

- remove version numbers within the kernel
- create seperate program context for each mining algorithm
- remove kernel `cn1_monero` is now integrated in `cn1`
- remname `cnX` kernel in `cnX + algorithmNumber`

fix #1218

- remove inline function with ugly macro :-(

Co-authored-by: psychocrypt <psychocryptHPC@gmail.com>
Co-authored-by: fireice-uk <fireice-uk@users.noreply.github.com>
Co-authored-by: Lee Clagett <code@leeclagett.com>
Co-authored-by: curie-kief <curie-kief@users.noreply.github.com>

fix double definition of define `MEM_CHUNK`

- add new option `comp_mode` to the amd config
- disable `if guards` within opencl kernel if `comp_mode : false`

allow usage of Mesa OpenCl

- add new option for `strided_index`
- add additional option if `strided_index == 2` to controll the memory chunk with

reduce usage of registers: based on the suggestion of @enerc77

remove branch differgences in AMD OpenCl code based on #454

a Please enter the commit message for your changes. Lines starting

Allow to change the indexing used to address the hash scratchpad memory.

- add option `strided_index` for each gpu

- add compile parameter to support aeon and xmr
- update auto suggestion to handle aeon

- remove 32bit target value
- use always 64bit target for amd backend

Early returns within a kernel can have unexpected behavior, this strongly depends on the opencl
runtime compiler. To avoid errors all early returns are removed.
There is no negative effect(e.g.performance) if the threads stay alive up to the end of the kernel.

One source for the early return is: http://al-key-opencl.blogspot.de/2014/09/be-careful-not-to-mix-early-return-and.html

This pull request also fix a race condition where all fill the shared memory.
Also a bug introduced with #16 is fixed, because of the early return not was possible that
the last block works with an wrong initilized shared memory (result should be a wrong hash if the result target size is valid).

Is is possible that the number of results are greater than the result output array,
in this case invalid memory can be access within the device and on the host side.

The number of threads within a kernel must be a multiple of the worksize.
If not it can crash on some systems.

- string literal can only be 16kbyte

- move source code to `src`
- categorize files and move to group folder
- change upper case class files to lower case
- change C++ header to `*.hpp`

- splitt literal to smaller chunks (windows only support 16k literals)
- remove copy command for opencl folder