chore(gatsby-plugin-benchmark-reporting): Add more meta data

[pvdz]

This adds a few new fields to the benchmark plugin: git repo hash, ci status, siteId, cwd, ci name, node version, gatsby version, gatsby cli version, sharp version (if present), webpack version, node memory (straight dump of process.memoryUsage()), size of all js files in root folder of public, and the number of images found recursively in public or .cache.

It also groups the versions and counts into a struct.

The plugin now reports the text of a 500 error, if any.

Need to update big query schema before pulling this. @jamo has updated the schema accordingly.

This change is not windows friendly due to the execSyncs using find, wc, bc, and forward slashes. There are probably one or two more subtleties that I can't test for. For the time being this plugin is intended to run with benchmarking service and some CI's, which all run on linux. If we need windows support, we can iterate (most-if-not-all of this stuff can be done with just JS, just more work).

Additionally the execs are sync because that just makes it easier. There are only a handful, and there is no performance concern because this runs at the end of a run. Changing it to async would not make a relevant difference but more annoying code to maintain.

$ BENCHMARK_REPORTING_URL=https://secret yarn bench
...
info Gathered data: {
  "time": "2020-02-13T12:21:21.207Z",
  "sessionId": "ed4fe1f3-afb3-4314-a493-f40715c3d5b9",
  "cwd": "/home/foo/gatsby/ifsc-150k",
  "timestamps": {
    "bootstrapTime": 1551,
    "instantiationTime": 1551,
    "benchmarkStart": 1594,
    "preInit": 1594,
    "preBootstrap": 1682,
    "preBuild": 2436,
    "postBuild": 7874,
    "benchmarkEnd": 7874
  },
  "gitHash": "cfa8e65c60b868bc5034b7c82f9330eb572dbc8c",
  "ci": false,
  "ciName": "local",
  "versions": {
    "nodejs": "v10.18.1",
    "gatsby": "2.19.16-dev-1581506531816",
    "gatsbyCli": "Gatsby CLI version: 2.8.29\nGatsby version: 2.19.16-dev-1581506531816\n  Note: this is the Gatsby version for the site at:/home/foo/gatsby/ifsc-150k",
    "sharp": "none",
    "webpack": ""
  },
  "counts": {
    "pages": 13,
    "jpgs": 0,
    "pngs": 0,
    "gifs": 0,
    "other": 0
  },
  "memory": {
    "rss": 322387968,
    "heapTotal": 247218176,
    "heapUsed": 157750664,
    "external": 151727365
  },
  "publicJsSize": 232929
}
info Flushing benchmark data to remote server...
info Server response: 201: Created

[pvdz]

(Requests for feedback were meant as a ping, but feedback is welcome. This cannot merge until db is updated to support the new fields.)

[pvdz]

Hm, should it also report the size of images? As a rough guide for triage of why image processing took so long or whatever?

[nicholascapo]

Still need CI_NAME

[pvdz]

@nicholascapo after some discussion we're pulling the siteId explicitly from the analytics object for now. If in the future more things get added to this env object it we can update accordingly to various implementations. Mainly wanted to prevent the footgun where the analytics object gets a new field while the benchmarking schema isn't expecting it and the server just refuses it.

[pvdz]

The siteId will be "error" if the tags value set and failed to parsed as json, just so we can track that just in case. It's empty if not set at all and should be the value supplied otherwise.