[logarchive] completely redesign the logarchive parser for performance

[cvandeplas]

• rewrite the parser completely (from scratch) , remove many layers and functions that make code flow difficult to read

• unifiedlogs_iterator stdout reading kill everything, CPU is only 25%.

• Let unifiedlogs save to a temp file, read the temp file while the process is still running and throw these lines into batches of lines to process

• write batches of records to the output file

• use orjson to improve performance loading and generating/saving json

• do some more tuning here and there

• work on unifiedlogs_iterator multithreaded, and initiate a PR to that project.