Unified memory allocations

[christiangnrd]

Close #601

@vchuravy The idea is that you request unified if you want to allocate some memory with unified memory to speed up operations that necessitate CPU access (like 1d reduce result or the indices container in findall). Otherwise just use the default storage type. Backends would have to add support for the unified kwarg.

[github-actions]

Your PR no longer requires formatting changes. Thank you for your contribution!

[vchuravy]

Perhaps we can make this backwards portable and add it to 0.9 first?

[vchuravy]

Thanks once we have it merged in 0.9, please update this PR so that the implementations match.

[codecov]

Codecov Report

❌ Patch coverage is 80.00000% with 4 lines in your changes missing coverage. Please review.
✅ Project coverage is 51.24%. Comparing base (1ac546f) to head (b0a1023).
⚠️ Report is 4 commits behind head on main.

Files with missing lines	Patch %	Lines
src/KernelAbstractions.jl	71.42%	4 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #630      +/-   ##
==========================================
+ Coverage   47.54%   51.24%   +3.70%     
==========================================
  Files          22       21       -1     
  Lines        1729     1610     -119     
==========================================
+ Hits          822      825       +3     
+ Misses        907      785     -122     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[vchuravy]

ready to merge?