BF: runscript - only consider statuses for success/failure

When running the usecase example ReproNimgh-438, the working tree is coming up
as unexpectedly dirty, including the change of status files from empty
to "succeed", apparently indicating that the content of the file is
not flushed at the time of the 'datalad add/save' call.  To prevent
saving before the content is flushed, don't count a status file unless
it has the content.

Modified-by: Kyle Meyer <[email protected]>
  Add commit message body, and dropped `find` call, as suggested by
  @yarikoptic.

Author: yarikoptic

reproman/support/jobs/job_templates/runscript/base.template.sh

@@ -53,7 +53,7 @@ then
 # least for now, below is a brittle solution were the last job waits until it
 # sees that all other jobs have exited and then runs the post-command stuff.
 nstatus () {
-    find "$metadir" -regex '.*/status\.[0-9][0-9]*' | wc -l
+    grep -E '^(succeed|fail)' "$metadir"/status.* | wc -l
 }
 
 # Ugly, but this sleep makes it less likely for the post-command tar to fail