[code-reflection] Integrated: Hat prep for cuda sync

[Gary Frost]

Lots of code changes here in prep for simplifying CUDA update. 

I decided to mirror the class heriarchy in ffi native code in the java FFIBackend.  

Also preparing the way for ffi backends to generate JSON perf data.

-------------

Commit messages:
 - Mirrored FFI hierarchy in FFIBackend.java
 - BuildInfo moved to CompilationUnit
 - Moved ffi code to better match CUDA/OpenCL common code
 - Add json parser/builder for perf counters
 - Templatize cl_xxx_info
 - Split the opencl backend native sources as a pattern for CUDA equivalents

Changes: https://git.openjdk.org/babylon/pull/370/files
  Webrev: https://webrevs.openjdk.org/?repo=babylon&pr=370&range=00
  Stats: 4879 lines in 45 files changed: 3630 ins; 1027 del; 222 mod
  Patch: https://git.openjdk.org/babylon/pull/370.diff
  Fetch: git fetch https://git.openjdk.org/babylon.git pull/370/head:pull/370

PR: https://git.openjdk.org/babylon/pull/370