Alpaka as 'git subtree' instead of 'git submodule'

[tdd11235813]

Changes Alpaka dependency workflow:

• replaces the git submodule with git subtree

• new workflow documented in README.md

• [x] merge after #104 (is required to work)

[tdd11235813]

• moved C++11 fix to #104 (so CPU backends might fail now, going to rebase after #104)
• git subtree workflow now documented in INSTALL.md (README.md is updated as well)
• Removes cupla_ALPAKA_PROVIDER
  • no internal git submodule updater, as git subtree is now used
  • only one find_package(alpaka)
• click on last commit to see that changes

Edit: version to 0.2.0 ?

[tdd11235813]

updated alpaka and rebased onto #104 to fix issue.

• cmake alpaka internal provider is back to provide explicit control of alpaka path selection

Note: rebasing with upstream lead git to be confused totally cuz of the submodule vs subtree (it mixed alpaka files with cupla files^^), re-adding subtree from scratch helped

[ax3l]

@tdd11235813 Pls note that we usually do not commit git subtree squashes under our own user name because it messes with the git statistics (e.g. you just authored 87k lines in cupla...). As an example, see: https://github.com/ComputationalRadiationPhysics/picongpu/pull/2893

[tdd11235813]

yeah @psychocoderHPC also mentioned this offline today. Any suggestions for the mail address? Not sure whether it needs to be a valid one. I would like to suggest a more generic one instead of cupla@hzdr.de, because it also will affect other [upcoming] (beta-)projects too, and I do not want to create myriads of addresses. Maybe crp-git@hzdr.de or something like that? I would add your workflow to the INSTALL.md then.

[psychocoderHPC]

cupla@hzdr.de is fine but let us move to the generic address crp-git@hzdr.de and use it in all new projects.

[ax3l]

Sure, please update your instructions in INSTALL.md with it :)

[tdd11235813]

ok, see PR #111