[Updated Draft] Lab

[ospira]

things to finish:

• better html layout (UI should highlight deopts, better block formatting, progressive display of content, flag structures that change v8 map for the same run - like during pnpm run micro arrays/deopt.js)
• mitata json format parser - for feature parity w the CLI output

things to note:

• traces like --trace-opt or --trace-deopt will write to the logFile somewhat unpredictably. thus, in order to capture these outputs within their proper context I adopt a delimiter model. much of the utils/ folder is simply creating good conventions around this to avoid parsing bugs.

• in this version i have commented out --trace-opt as i opted for traceLogger to instead for an optimization after the first bench in a set/catalog runs: https://github.com/pmndrs/koota/pull/168/files#diff-79e44628c4e8559e0e0592c1aa9b829df446fecc84d5659a4425ae99bef438ddR39-R43

the thought process for this is what we really want to flag here is deopts. we could wait for the function to manually become hot and stable, but we wouldn't necessarily know when this would happen. and it doesn't seem like --trace-opt gives detailed output anyway - most of the logs from that will just give you the memory address and the only "reason" given is the function is now "hot & stable". i could have the html handle present these in a useful way - or at least a way that wasn't annoying - but really they are not that meaningful imo compared to a deopt.

• a second pass of the HTML could do things like request more context from the logFile - like get bytecode. the json format outputted by mitata is extremely verbose. i have a wip parser for this as well (things to finish ✨ core: create AoS trait stores #2). currently the mitata runner is set to output using the markdown format option - which is really just a placeholder. The JSON parser will aim to have feature parity with mitata's format for CLI output.