Memory improvements by only returning the needed JSON output via NodeJS

[mattiasgeniar]

The NodeJS wrapper returned the entire JSON output of the Lighthouse check, which contains 3 major blocks:

• lhr
• artifacts
• report

We only need the HTML & JSON output that is in report and any runtime errors in lhr, the rest currently isn't used. This PR filters only the necessary fields in NodeJS and prevents PHP from re-encoding the JSON string, which in extreme cases causes out-of-memory issues.

This PR remains simple and if there's ever a need for more data out of the Lighthouse package, it should be simple to extend the NodeJS output.

Peak memory usage results from testing:

URL	Before	After	Change Percentage
https://ohdear.app	1221MB	13.2MB	-98.92%
https://example.com	17MB	9MB	-47.06%
https://spatie.be	63MB	10.1MB	-83.97%
-Redacted-	415MB	12.39MB	-97.01%

Even in simple HTML examples the gains are ~50%, in edge-cases (ie complex sites) it can be as high is 90%+.

[freekmurze]

Great work, I didn't realize these reports were so big.

[freekmurze]

Thanks!