Use object rather than array for scores.json entries

[nicoburns]

This changes the scores.json format in preparation for #50 (rewrite score calculation in Rust). The main change is to replace the use of a heterogeneous array (mix of strings and objects) with a keyed object and a nested array. The new format is easier to serialize/deserialize using serde (and IMO makes more sense in JS anyway).

Corresponding servo.org PR (should be merged at same time): servo/servo.org#241

before:

[
    "2025-07-17",
    "1c4797809",
    "0.0.1-f70a4eb4f",
    {
      "total_tests": 46787,
      "total_score": 29447.8764608724,
      "total_subtests": 1854463,
      "total_subtests_passed": 1722785
    },
    {
      "total_tests": 809,
      "total_score": 502.872027417027,
      "total_subtests": 3659,
      "total_subtests_passed": 2186
    },
    ...
]

After:

{
        "date": "2025-07-17",
        "wpt_revision": "1c4797809",
        "servo_revision": "0.0.1-f70a4eb4f",
        "scores":
        [
            {
                "total_tests": 46787,
                "total_score": 29447.876460872423,
                "total_subtests": 1854463,
                "total_subtests_passed": 1722785
            },
            {
                "total_tests": 809,
                "total_score": 502.8720274170274,
                "total_subtests": 3659,
                "total_subtests_passed": 2186
            },
            ...
        ]
}

[mukilan]

I'm curious to know what is the new size of the scores.json. Hopefully, the gzip compression on server side still helps.

[nicoburns]

I'm curious to know what is the new size of the scores.json

The new file is 2147kb compared to 2107kb for the old file. GZIP compression is reducing the old files to 76.7kb, and I'd expect the new changes to also compress well as they're adding a single repeated string.

[nicoburns]

New gzipped size is 77.3kb