(PUP-11685) Update total fact count

[mhashizume]

Puppet calculates the total number of facts on an agent node to warn the user if it will cause performance issues with PuppetDB. PuppetDB performance can be impacted when the leaf count of a factset is too high.

Prior to this commit, Puppet inconsistently calculated facts when generating a warning. This did not count individual array elements; for example, the processor.models fact with a fact per processor core would only be counted once. Additionally, hash keys were incorrectly each counted as individual facts when they would not represent separate leaf counts in PuppetDB.

These behaviors lead to a slight discrepancy between what Puppet would report to the user and what the total fact count ended up being in PuppetDB.

This commit updates Puppet to more accurately count facts to reflect a factset's leaf count in PuppetDB.

[mhashizume]

Here are some examples from my testing for this PR.

This is Puppet's current behavior when counting total facts (using Puppet 7.27.0 with legacy facts omitted):

root@ethical-stuff:~# puppet agent -t --debug | grep 'The total number of facts registered is'
Debug: The total number of facts registered is 360

This is the updated behavior using the code in this PR:

root@ethical-stuff:~# puppet agent -t --debug | grep 'The total number of facts registered is'
Debug: The total number of facts registered is 352

This is what a PuppetDB query returns:

root@younger-crepe:~# curl -sX GET http://localhost:8080/pdb/query/v4/fact-contents --data-urlencode 'query=["=", "certname", "ethical-stuff.delivery.puppetlabs.net"]' | jq '. | length'
356

The updated behavior is more accurate to what is reported to PuppetDB. The difference of four facts between the two counts is due to the presence of trusted facts in PuppetDB.

Overall while this new count is more accurate, it seems like generally there won't be a massive difference from the old behavior. The old methodology overcounted some things (hash keys) but undercounted others (array elements), so it seems like it overall kind of balanced out.

[mhashizume]

@Sharpie @rbrw thanks for your earlier feedback on this issue. Would you be able to review this when you have the chance?

Thank you!