The current tap13 support summarizes testcases on a rather coarse
granularity. Which sort of defeats the purpose, especially in CI
environments, where having results for individual tests to pinpoint
problems helps a lot.

To improve things, this patch makes TAP mode bypass print_result and extract
results for individual testcases as they are sent to the log file.  It then
introduces test case numbers and writes the plan at the end.

One could argue that post-processing the logs is a rather fragile approach.
Which is true - but apparently already the case, see e.g. extract_summary
in scripts/runtime.bash. Plus this is quite cheap, while a proper solution
would require to modify the kernels, so we can e.g. pass in a counter for
the test case number. And we would probably have to come up with reasonable
test case names, while this approach simply derives them from each test
output.

Before:
  $ ./run_tests.sh -t
  TAP version 13
  ok selftest-setup
  ok intercept
  ok emulator
  ok sieve
  ok sthyi
  ok skey
  ok diag10
  ok pfmf
  ok cmm
  ok vector
  ok gs # SKIP
  ok iep # SKIP
  1..12

After:
  $ ./run_tests.sh -t
  TAP version 13
  ok 1 - cmm: privileged: Program interrupt: expected(2)
  ok 2 - cmm: invalid ORC 8: Program interrupt: expected(6)
  ok 3 - diag10: lowcore freeing: 0x0000/0x0000: Program interrupt: expected(6)
  ok 4 - diag10: lowcore freeing: 0x1000/0x1000: Program interrupt: expected(6)
  ok 5 - diag10: lowcore freeing: 0x0000/0x1000: Program interrupt: expected(6)
  ok 6 - diag10: start/end: end < start: Program interrupt: expected(6)
  ok 7 - diag10: start/end: unaligned start: Program interrupt: expected(6)
  ok 8 - diag10: start/end: unaligned end: Program interrupt: expected(6)
  ok 9 - diag10: privileged: Program interrupt: expected(2)
  ok 10 - emulator: spm/ipm: cc=0,key=f: bit 32 and 33 set to zero
  ok 11 - emulator: spm/ipm: cc=0,key=f: bit 0-31, 40-63 unchanged
  ok 12 - emulator: spm/ipm: cc=0,key=f: cc and key applied
  ok 13 - emulator: spm/ipm: cc=1,key=9: bit 32 and 33 set to zero
  1..180
  [...]

Signed-off-by: Stefan Raspl <raspl@linux.ibm.com>
[Almost rewritten, but still giving a lot of credit to Stefan. - Paolo]
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>