feat(bench): publish ESP32-S3 build timings alongside Arduino Uno #1488

Context

The nightly build-comparison workflow currently benchmarks only the Arduino Uno Blink fixture. Its published benchmark.svg heading/description name Arduino Uno, and latest.json exposes one result set containing Arduino CLI, PlatformIO, and fbuild timings.

Consumers need the same cold/warm comparison for ESP32-S3 in the existing stable artifact at:

https://raw.githubusercontent.com/FastLED/fbuild/benchmark-stats/benchmark.svg

Proposal

Extend the existing Blink benchmark with an ESP32-S3 PlatformIO environment and make the benchmark runner describe a board by its display name, Arduino FQBN, PlatformIO/fbuild environment, and toolchain pins. Run the same three-tool cold/warm measurement for Arduino Uno and ESP32-S3, then render both board groups into the existing SVG and machine-readable artifacts.

Keep one workflow, one fixture, one benchmark runner, and one published benchmark-stats branch.

Acceptance criteria

• RED: add a focused runner test proving the current SVG/output model cannot represent both Arduino Uno and ESP32-S3 result groups.
• GREEN: the focused benchmark-runner tests pass after the output model and renderer support both boards.
• bench/blink/platformio.ini includes an ESP32-S3 environment usable by PlatformIO and fbuild.
• The workflow installs/primes both Arduino Uno and ESP32-S3 toolchains outside timed measurements.
• Each board is measured with Arduino CLI, PlatformIO, and fbuild using the same cold/warm trial policy.
• The existing benchmark.svg contains two labeled board groups: Arduino Uno and ESP32-S3.
• The machine-readable latest/history output preserves board identity for both groups.
• After the workflow publishes, curl -fsSL https://raw.githubusercontent.com/FastLED/fbuild/benchmark-stats/benchmark.svg returns an SVG containing both Arduino Uno and ESP32-S3.

Decisions

• Use the existing Blink fixture and runner because this is the exact comparison path already published for Uno; duplicating a benchmark implementation would invite drift.
• Keep the existing artifact URL because downstream curl/image consumers should gain the second board without discovering a new endpoint.
• Benchmark the same three tools for both boards so the two graph members remain directly comparable.
• Treat this as a benchmark schema extension rather than a second standalone report so board identity is explicit in SVG, latest JSON, and history.