Sort plot-schema and add test to track plot-schema changes

[archmoj]

Resolves #5588.
In PRs after this when there is a change in attributes (e.g. to implement a new feature), one should simply run

npm run schema

and commit the test/plot-schema.json file.

On CircleCI, that schema file would be tested against the auto-generated baseline at dist/plot-schema.json.
In the case of any difference a diff is presented and test should fail.
We also added a message on CircleCI asking contributors to run npm run schema and commit the results in the "test" folder (i.e. similar to baseline changes).

This would help track the changes to the API way before release cycles, right when a feature/documentation PR is submitted or even before.

In addition to above,

1. As a result of this PR the plot-schema object is sorted to be as stable as possible for future comparisons.
2. The compress_attributes browserify transform is now disabled to generate plotly.js files in the build folder which not only decreases the build & rebuild times but also help make this possible.

@plotly/plotly_js

[alexcjohnson]

Actually we need to dig into arrays too, right? We've got stuff in the schema like:

      "items": [
       {
        "valType": "number",
        "min": 0,
        "max": 1,
        "editType": "calc"
       },
       {
        "valType": "number",
        "min": 0,
        "max": 1,
        "editType": "calc"
       }
      ],

[archmoj]

Good call.
aea3022 also fixed cases where arrays e.g. enumerated valType turned into objects!
Before:

   "values": {
    "0": "forward",
    "1": "reverse"
   }

After

   "values": [
    "forward",
    "reverse"
   ]

[alexcjohnson]

💃