Clean up the x-mas tree

unroll AES round by hand to help the compiler to optimize the command queue

Change the default for strided index. Th euser feedback shows that strided_index = true will give a view hashes more per second.

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

- add option `strided_index` for each gpu

Fix for AMD GPU detection on MacOS X
On MacOS X the platform is reported as Apple, and the vendor is reported as AMD.

Fixing the compile error on MacOS X

The max work size shown to the user was the hardware maximum but not take in account thet some kernel spwning 8 times more threads per work group than configured by the user.

- handle opencl errors to avoid unexpected behaviors
- remove code duplicatio nwithin the some method

- enforce that only host side OpenCl 1.2 functions are used if the miner is compiled with `XMR-STAK_COMPILE=generic`
- update documentation
- add CMake option `XMR-STAK_COMPILE` to windows

Remove whitespace lines

- removed all space indention in `*.hpp` and `*.cpp` files

* CONF_NO_TLS fix

* gpu error fix + always return to sched after pin

Pass threadno up the food chain

- add type of the backend to each backend-plugin
- add `gteName` to `iBackend` to get the name of the backend

- allow to use up to 2024 threads on VEGA gpus

fix #5

- increase safty memory from 64 to 128 MiB
- NVIDIA: increase lmem reserve per thread to 1kiB

- rename all `xmr` to `monero`
- be insensitive while check for set currency
- add function to compate two strings insensitive

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

- check if affinity vale before calling `thd_setaffinity`

This is a follow up of #43 and use the some mechanism to set the thread affinity for
non cpu backends correct.

- use cpu affinity workflow for nvidia and amd
- cpu: move messages of thread spawning before thread creation

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

fix bug announced in #52

If the miner is compiled in debug mode the assert condition results in a compile time error.

- use always a multiple of the number of compute units
- add more device information to the config file

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.

check that the number of global threads is a multiple of the workgroup size

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

Platform index zero was always used to search for the device detection.

Fix: use correct detected AMD platform index

use nicehash start nonce in all backends if nicehash is enabled

- string literal can only be 16kbyte

- add clang compile section
- cleanup CMake file
- fix missing return value