Fix an error caused by system's /tmp being full when running checks in lbnl_hw.nhc

[RJukkara]

The error in question can be easily reproduced by having for example check_hw_cpuinfo, check_hw_eth or check_hw_physmem configured in the system's nhc.conf and filling up /tmp.
Currently what happens is:

[root@c1104 ~]# df -h /tmp/
Filesystem      Size  Used Avail Use% Mounted on
tmpfs           512M  512M     0 100% /tmp
[root@c1104 ~]# nhc -ad
DEBUG:  Debugging activated via -d option.
ERROR:  nhc:  Health check failed:  check_hw_cpuinfo:  Actual CPU socket count (1) does not match expected (2).
ERROR:  nhc:  Health check failed:  check_hw_physmem:  Actual RAM size (0 kB) less than minimum allowed (263504064 kB).
ERROR:  nhc:  Health check failed:  check_hw_eth:  Ethernet device ib0 not detected.
ERROR:  nhc:  3 health checks failed.

After the changes in the PR:

[root@c1104 ~]# df -h /tmp/
Filesystem      Size  Used Avail Use% Mounted on
tmpfs           512M  512M     0 100% /tmp
[root@c1104 ~]# nhc -ad
DEBUG:  Debugging activated via -d option.
[root@c1104 ~]# echo $?
0