Cache preformatted tokens in statistics task

[mutatrum]

Follow-up of #2016. This change amortises the serialisation time of all values at stats collection time. When the front-end then requests a full 720 point snapshot, the serializer just dumps the strings. This reduced lock time on the statistics object, reduces TTFB and response time.

	Stock cJSON	sprintf	yyjson	pre-formatted
Time to First Byte (TTFB)	~316 ms	~36 ms	~61 ms	~25 ms
Total Response Time	~357 - 400 ms	~61 - 75 ms	~105 - 131 ms	~76 – 103 m
Transferred Payload Size	56.8 KB	38.0 KB	37.8 KB	38.0 KB

It also makes the code simpler, http_server no longer has knowledge of which fields there are and several lists have been eliminated.

[github-actions]

Test Results

  2 files  ±0    2 suites  ±0   1s ⏱️ -1s
252 tests ±0  252 ✅ ±0  0 💤 ±0  0 ❌ ±0 
254 runs  ±0  254 ✅ ±0  0 💤 ±0  0 ❌ ±0 

Results for commit 06de123. ± Comparison against base commit 405a7bf.

[0xf0xx0]

tested ack