selftests/bpf: Add parallelism to test_progs

This patch adds "-j" mode to test_progs, executing tests in multiple process.
"-j" mode is optional, and works with all existing test selection mechanism, as
well as "-v", "-l" etc.

In "-j" mode, main process use UDS/DGRAM to communicate to each forked worker,
commanding it to run tests and collect logs. After all tests are finished, a
summary is printed. main process use multiple competing threads to dispatch
work to worker, trying to keep them all busy.

Example output:

  > ./test_progs -n 15-20 -j
  [    8.584709] bpf_testmod: loading out-of-tree module taints kernel.
  Launching 2 workers.
  [0]: Running test 15.
  [1]: Running test 16.
  [1]: Running test 17.
  [1]: Running test 18.
  [1]: Running test 19.
  [1]: Running test 20.
  [1]: worker exit.
  [0]: worker exit.
  #15 btf_dump:OK
  #16 btf_endian:OK
  #17 btf_map_in_map:OK
  #18 btf_module:OK
  #19 btf_skc_cls_ingress:OK
  #20 btf_split:OK
  Summary: 6/20 PASSED, 0 SKIPPED, 0 FAILED

Know issue:

Some tests fail when running concurrently, later patch will either
fix the test or pin them to worker 0.

Signed-off-by: Yucong Sun <sunyucong@gmail.com>

V4 -> V3: address style warnings.
V3 -> V2: fix missing outputs in commit messages.
V2 -> V1: switch to UDS client/server model.

Author: thefallentree