[x-pack][metricbeat][iis] Investigate metrics leaking to a wrong process (application pools)

It seems that when you run multiple application pools and one of them gets stopped/restarted metricbeat reports its metrics (for example `# Gen 0 Collections`) as metrics of a different app pool (that is still running).

### Preliminary analysis
It seems this behavior is connected to how PDH counters behave.
In order to try and investigate this a little further I've used this powershell script:
```
$results = foreach ($proc in Get-CimInstance Win32_Process -Filter "Name = 'w3wp.exe'") {
$WorkerPID = $proc.ProcessId
$appPoolName = $proc.CommandLine.Split('"')[1]

# Find the instance name (w3wp, w3wp#1, etc)
$processCategory = New-Object System.Diagnostics.PerformanceCounterCategory("Process")
$instances = $processCategory.GetInstanceNames() | Where-Object { $_ -match "^w3wp(#\d+)?$" }

foreach ($inst in $instances) {
$idCounter = New-Object System.Diagnostics.PerformanceCounter("Process", "ID Process", $inst, $true)

# Check if this instance belongs to our current loop's PID
if ($idCounter.RawValue -eq $WorkerPID) {
try {
# Create the Gen 0 counter
$gen0 = New-Object System.Diagnostics.PerformanceCounter(".NET CLR Memory", "# Gen 0 Collections", $inst, $true)

# First call to NextValue() often returns 0, we need the RawValue or a second call
$val = $gen0.RawValue

[PSCustomObject]@{
AppPool = $appPoolName
PID = $WorkerPID
Instance = $inst
Gen0Cols = $val
}
} catch {
# This w3wp isn't running .NET Framework (could be .NET Core or No Managed Code)
}
break # Found the instance for this PID, move to next process
}
}
}

$results | Format-Table -AutoSize
```

This script produced a table like this:
```
AppPool PID Instance Gen0Cols
------- --- -------- --------
DefaultAppPool 1404 w3wp 3
zRmAoVOT 12932 w3wp#1 3
.Ney_2 7296 w3wp#2 3
EtQKqWjM 624 w3wp#3 3
lQzBAMHP 5364 w3wp#4 11
ytYuhpwr 11660 w3wp#5 3
zOXGgNtn 2044 w3wp#6 4
lgArfpvj 6972 w3wp#7 3
cFiwlXuz 11236 w3wp#8 3
BwFRIWmh 10244 w3wp#9 3
qixghmay 12536 w3wp#10 3
mxIgZapB 5288 w3wp#11 4
bklVLKgi 10576 w3wp#12 4
uQlBwkFb 3108 w3wp#13 3
rJYpZcHu 8196 w3wp#14 3
qRIByVab 3300 w3wp#15 3
yFTULeYC 4780 w3wp#16 4
DkYZNoMP 11232 w3wp#17 4
PbrEIdkD 12144 w3wp#18 3
AjrdVzqD 9616 w3wp#19 4
tzJjKvsN 8948 w3wp#20 0
HbniZJXy 7912 w3wp#21 18
Classic .NET AppPool 8664 w3wp#22
```
After that I've stopped `lQzBAMHP` app pool(Instance `w3wp#4`, Gen0Cols: 11) and ran the script again and got following result:
```
AppPool PID Instance Gen0Cols
------- --- -------- --------
DefaultAppPool 1404 w3wp 3
zRmAoVOT 12932 w3wp#1 3
.Ney_2 7296 w3wp#2 3
EtQKqWjM 624 w3wp#3 3
ytYuhpwr 11660 w3wp#4 11
zOXGgNtn 2044 w3wp#5 3
lgArfpvj 6972 w3wp#6 4
cFiwlXuz 11236 w3wp#7 3
BwFRIWmh 10244 w3wp#8 3
qixghmay 12536 w3wp#9 3
mxIgZapB 5288 w3wp#10 3
bklVLKgi 10576 w3wp#11 4
uQlBwkFb 3108 w3wp#12 4
rJYpZcHu 8196 w3wp#13 3
qRIByVab 3300 w3wp#14 3
yFTULeYC 4780 w3wp#15 3
DkYZNoMP 11232 w3wp#16 4
PbrEIdkD 12144 w3wp#17 4
AjrdVzqD 9616 w3wp#18 5
tzJjKvsN 8948 w3wp#19 0
HbniZJXy 7912 w3wp#20 18
Classic .NET AppPool 8664 w3wp#21
```
Notice that with instance w3wp#4 and Gen0Cols of 11 now a different app pool is reported (`ytYuhpwr`). And metricbeat has exactly same behavior.

This gh issue is meant for:
1. Investigation of why exactly this behavior happens (likely it's something about how courters work but it has be confirmed).
2. Find a solution for metricbeat to make metrics reporting correct in an event of app pool being stopped/restarted.