initrd: use nm_utils_strsplit_set_with_empty() to split strings in nmi_dt_reader_parse()

strsplit() can be done in O(n) runtime and one memory allocation.
g_strsplit() creates a deepcopied strv array, requiring allocate
memory for each token separately. Also, it's implemented by constructing
the list with a tempory GSList, so it actually needs two allocations
for each tokens.

I think we should not use such wasteful API, if we don't need a clone
of the tokens.
18 jobs for th/initrd-parse-cleanup in 36 minutes and 2 seconds (queued for 21 seconds)
Status Job ID Name Coverage
  Test
passed #1067739
checkpatch

00:01:24

manual #1067745
allowed to fail manual
t_centos:7.5.1804
manual #1067746
allowed to fail manual
t_centos:7.6.1810
manual #1067752
allowed to fail manual
t_debian:10
manual #1067751
allowed to fail manual
t_debian:9
manual #1067754
allowed to fail manual
t_debian:sid
manual #1067753
allowed to fail manual
t_debian:testing
manual #1067740
allowed to fail manual
t_fedora:28
manual #1067741
allowed to fail manual
t_fedora:29
passed #1067742
t_fedora:30

00:34:38

manual #1067743
allowed to fail manual
t_fedora:31
manual #1067744
allowed to fail manual
t_fedora:rawhide
manual #1067747
allowed to fail manual
t_ubuntu:16.04
manual #1067748
allowed to fail manual
t_ubuntu:18.04
manual #1067750
allowed to fail manual
t_ubuntu:devel
manual #1067749
allowed to fail manual
t_ubuntu:rolling
 
  External
failed https://desktopqe-jenkins.rhev-ci-vms.eng.rdu2.redhat.com/job/beaker-NetworkManager-gitlab-trigger-code-upstream/943/

failed https://desktopqe-jenkins.rhev-ci-vms.eng.rdu2.redhat.com/job/beaker-NetworkManager-gitlab-trigger-code-upstream/943/

05:06:20