[RFC PATCH v2 0/2] Imporove remote helper exec failure reporting

(This is newer than updated patch v3).

I decided to go on a bit further with new features, and that's why this is back as RFC.

Diffrences from first RFC round:

- Split loop-over-close to its own function and warn but not retry on
  bizarre error codes (basically anything except EBADF and EINTR).
- Don't rely on pipe(2) to preserve the fd array on failure.
- Add some testcases.
- Fix buggy behaviour if report pipe read end winds up as fd #0.
- Set silent_exec_failure on executing remote helper.
- Make exec reporting in presence of preexec_cb controlled by flag to
  solve pager deadlock problems without preventing using preexec_cb
  with exec reporting entierely (the way pager is invoked is just
  fundamentially incompatible with exec reporting).
- Add flag to make run-command give different code for exec failing due to
  ENOENT than for other failures (including other ways to get ENOENT).
- change chdir failure to be reported as error (and pass error code back).
  BTW: Nothing seems to use that chdir feature, aside of new testcase.
- Report errors to parent's stderr.
- Don't retry waitpid on bizarre error codes (basically anything except
  EINTR).
- Make run_command actually complain about exec failing (as API docs
  say).

Ilari Liusvaara (2):
  Report exec errors from run-command
  Improve transport helper exec failure reporting

 Documentation/technical/api-run-command.txt |    2 +
 Makefile                                    |    1 +
 run-command.c                               |  196 +++++++++++++++++++++++++--
 run-command.h                               |    2 +
 t/t0061-run-command.sh                      |   20 +++
 test-run-command.c                          |   74 ++++++++++
 transport-helper.c                          |   27 +++-
 7 files changed, 305 insertions(+), 17 deletions(-)
 create mode 100755 t/t0061-run-command.sh
 create mode 100644 test-run-command.c