New Disk Monitor, stale truncate file fix

1. when crashing, mysql sometimes leaves behind truncate files. these mess up restarts
   after replication, since they will contain stale table state. they are now removed.

2. the disk monitor had an unfortunate sensitivity to version, where the columns displayed
   varied between OS releases. now, it is much smarter, parsing the heading and associating
   the headings with metric names, and then binding the corresponding columns to metrics
   at run time. this no longer requires modification based on column count; it does the
   right thing when columns are added or deleted, or even renamed. if a new version decides
   to rename a column, then it can be added to the translation table and things will just
   work.

parallel rsync errors to STDERR and log file

parallel replicate rsync thread errors now copied to STDERR and saved to log file (previously only emitted to STDERR)

OPENSSL indirection

if an OPENSSL variable is set, use it

Avoid copying datadir twice if within <controller install>/data etc

Fix to prevent rsync of controller install dir also wrongly copying datadir if within $APPD_ROOT/data or some such sub-directory. Duplicate copying of datadir leads to replication failure with "master and slave have equal MySQL server UUIDs".

disable /proc/sys/kernel/numa_balancing

disabled /proc/sys/kernel/numa_balancing and other small changes

Preventing chatty boks_ssh installations breaking replicate.sh

Minor patch to prevent chatty ssh installations (e.g. boks_ssh) from breaking replicate.sh.

Mds_topology rm fix

1. the get_privilege_escalation has a path where ssh sudo is run. this requires that ssh be run with the
   -t option to allocate a pty.

2. replicate had a too-wide rm -f logs/log, intending to prune bin-logs and relay-logs. this also hit the mds_topology
   database directory.

Warn if Linux install pre-requisites have not been completed

1. Added warning message (that will not fail the replicate.sh) if either 'max processes' or 'max open files'
   are below values required within the Controller's Linux installation pre-requisites. Can save many hours
   of tracking down weird process failures.

Fix to permit spaces in Linux group names

1. Fixed failure in replicate.sh if Linux group name of AppD user contained a space e.g. 'domain users'

2. Fixed rare failure in replicate.sh if file to be rsync'd has been deleted by time rsync starts.
   Just the return code, 24, for vanished input file has been added to rsync's ignore list.

Avoiding false replicate discrepancies due to differing server LOCALE's

1. Added explicit LOCALE setting for replicate sort calls to avoid rare cases that wrongly claim
   differences/discrepencies in files between primary and secondary. Having HA servers with different
   LOCALE's causes differences in sort order that lead to wasted effort with too many files being
   transferred again. This issue does not affect correctness of the replication setup but can slow
   the setup process down.