perfdata: add experimental generic collector

[jkroepke]

Ref #1350
Fixes #1414

This PR setup a generic performance data collectors based on the users configuration.

Instead existing collectors that is using the perflib, the perfdata collector using the Performance Data Helpers function. See #1350

Implement the Performance Data Helpers as separate collector may have multiple oppunities:

It's easier to compare data between Performance Data Helpers and Performance Data from perflib

Any PerfData custom metrics can be implement now, without having the hard requirement of implement collector of each single subsystem.

Example:

--collectors.enabled="perfdata" --collector.perfdata.objects='[
  {"object":"Processor Information","instances":["*"],"counters": {"% Processor Time": {}}},
  {"object":"Memory","counters": {"Cache Faults/sec": {"type": "counter"}}}
]'

Produces this:

Details

# HELP windows_perfdata_memory_cache_faults_sec 
# TYPE windows_perfdata_memory_cache_faults_sec counter
windows_perfdata_memory_cache_faults_sec 2.3861366e+07
# HELP windows_perfdata_processor_information__processor_time 
# TYPE windows_perfdata_processor_information__processor_time gauge
windows_perfdata_processor_information__processor_time{instance="0,0"} 1.7665375e+11
windows_perfdata_processor_information__processor_time{instance="0,1"} 1.797465625e+11
windows_perfdata_processor_information__processor_time{instance="0,10"} 2.315621875e+11
windows_perfdata_processor_information__processor_time{instance="0,11"} 2.3535375e+11
windows_perfdata_processor_information__processor_time{instance="0,12"} 2.37738125e+11
windows_perfdata_processor_information__processor_time{instance="0,13"} 2.3755578125e+11
windows_perfdata_processor_information__processor_time{instance="0,14"} 2.3753984375e+11
windows_perfdata_processor_information__processor_time{instance="0,15"} 2.3741625e+11
windows_perfdata_processor_information__processor_time{instance="0,16"} 2.3767484375e+11
windows_perfdata_processor_information__processor_time{instance="0,17"} 2.3750421875e+11
windows_perfdata_processor_information__processor_time{instance="0,18"} 2.3664765625e+11
windows_perfdata_processor_information__processor_time{instance="0,19"} 2.370809375e+11
windows_perfdata_processor_information__processor_time{instance="0,2"} 1.997065625e+11
windows_perfdata_processor_information__processor_time{instance="0,20"} 2.373096875e+11
windows_perfdata_processor_information__processor_time{instance="0,21"} 2.3738828125e+11
windows_perfdata_processor_information__processor_time{instance="0,22"} 2.3773328125e+11
windows_perfdata_processor_information__processor_time{instance="0,23"} 2.37351875e+11
windows_perfdata_processor_information__processor_time{instance="0,3"} 1.9896265625e+11
windows_perfdata_processor_information__processor_time{instance="0,4"} 2.2467421875e+11
windows_perfdata_processor_information__processor_time{instance="0,5"} 2.3189421875e+11
windows_perfdata_processor_information__processor_time{instance="0,6"} 2.301384375e+11
windows_perfdata_processor_information__processor_time{instance="0,7"} 2.3463625e+11
windows_perfdata_processor_information__processor_time{instance="0,8"} 2.3394984375e+11
windows_perfdata_processor_information__processor_time{instance="0,9"} 2.360415625e+11

In case it works well, some collectors will be migrate from perflib to PDH, too.

[DerfOh]

Any PerfData custom metrics can be implement now, without having the hard requirement of implement collector of each single subsystem.

Seems this might resolve #1414 as well

[breed808]

Excellent work on getting this implemented! I've left a few minor questions, let me know if they require any clarification.