Searching for Poller Items Generates SQL Errors

* On large systems statistics gathering at the beginning of a poller run lead to excessive polling times
* It's important to remove all items from the poller_output table for the poller as it impacts the end of the previous poller
* The other change here has to do with large databases where the poller_output memory table can get pushed into swap.
* Moved statistics out of the internal loop, and made a few optimizations for the $totals query to perform better.
* We only care about total_ports if the snmp_port is > 0

This is a compromise with regard to a possible issue with MariaDB 10.3 to only allow this via setting, and only allow for single poller systems.

We will look to re-design this in a future Cacti release.

This test added multi-output values that return a NaN or 'U'.

Edit Graph Template link missing base_url.

Template Export missing Graph Template columns multiple and test_source

* This change will add additional security to prevent certain classes of exploits in Cacti's use of the unserialize function.
* Additionally, adding an additional setting to dsv_log() to reduce the amount of logging during use of the tool to test data sources vs. poller operations.

* SNMP-Options - Bulk Walk Maximum Repetitions ignored during save
* light PSR to remove tabs and replace with spaces.

"Balance Process Load" dont work after upgrade to 1.2.23

* Rebuilding Poller Cache from Utilities does not respect poller interval due to lack or ordering
* Decrementing of rrd_next_step is flawed

* We still have references in the code to 'poller_lastrun' without the poller id.  Set the legacy value if the poller_id is 1
* If the poller_interval equals the cron interval, you can still have to bypass the scalability enhancement if you largest data source profile is larger than the poller interval