Make timestamp formatting consistent

[amousset]

Use %jd + intmax_t to format the timestamp (like other time_t formats do) in file backup code.

On ARMv7 (which is, understandably, unsupported) it segfaults at runtime. But formatting a (signed) time_t as unsigned is, AFAIK, undefined behavior annyway.

[craigcomstock]

Seems reasonable. Thanks for the contribution. Hope your ARMv7 systems are doing well.

[amousset]

Fixed other similar warnings (although less serious) with the same change for consistency.

locks.c:356:79: warning: format '%ld' expects argument of type 'long int', but argument 4 has type 'time_t' {aka 'long long int'} [-Wformat=]
  356 |         Log(LOG_LEVEL_DEBUG, "Giving time to process '%d' (holding lock for %ld s)", entry->pid, (now - entry->time));
      |                                                                             ~~^                  ~~~~~~~~~~~~~~~~~~~
      |                                                                               |                       |
      |                                                                               long int                time_t {aka long long int}
      |                                                                             %lld

[larsewi]

@cf-bottom Jenkins please :)

[cf-bottom]

Alright, I triggered a build:

Jenkins: https://ci.cfengine.com/job/pr-pipeline/13378/

Packages: http://buildcache.cfengine.com/packages/testing-pr/jenkins-pr-pipeline-13378/

[craigcomstock]

sequential-tests failure was a flake, looks good.

[craigcomstock]

@amousset please squash all commits into one or separate into issue fix and code quality fix.

[amousset]

Squashed.