fluentbit/processor: Update the parser to reserve data while avoiding duplication of keys

[ridwanmsharif]

Fixes b/226372668.

This change does the following:

• Adds a nest LUA component to nest all fields (except the parse key) under a temporary internal field
• Parse the payload using the parse key while using Reserve_data set to "True"
• Adds a merge LUA component to merge the parsed payload with the original. It will overwrite all fields except for the logName (which will be preserved from the original) and the labels which will be merged with the parsed payload. See details below. Then delete the temporary internal field.
• The above is now applied to all parsers used in the Ops Agent
• Update the mongodb parser to use the same logic as above
• Integration test for the above changes

Logic for merging labels:

• All labels in the original payload are kept as long as the parsed payload doesn't conflict with the same key
• If a parsed payload has a label with the same key as the original, then it overwrites the original
• All labels in the parsed payload are kept

See full results for the perf benchmarks here: https://screenshot.googleplex.com/5sZHUPnyjMvDXnz

To make sense of the results use the following legend: file -> file receiver with no parsing file_parse_raw -> file + parser file_parse -> file + parse with some fields that exist before parsing (added using another filter thats why there is some perf impact) file_parse_lua -> file + parse + some existing fields + 2 lua filters to deal with deduplication (using the slowest path: some original fields already present and needs merging)

[qingling128]

LGTM

[ridwanmsharif]

The integration test added in the PR passes on Windows. The only failures are unrelated (and a known issue) and are being addressed in b/235115328. Merging this PR now.