[draft] feat(monitoring): add byte metrics with aggregate Prometheus gauges

[gimballock]

Add bytes_received/bytes_sent tracking across all three components
(pool, jd-client, translator) for SV2 channels and SV1 clients.

SV2 byte tracking:

• Per-channel (bytes_received, bytes_sent) counters in pool, jd-client,
  and translator, keyed by channel_id in bytes_by_channel maps
• Shared mining_message_channel_id() helper in protocol_message_type.rs
  to extract channel_id from parsed Mining messages
• Raw frame payload guard (len >= 4) for channel_id extraction from
  unparsed upstream frames in jd-client and translator
• Cleanup in all CloseChannel handlers and ChannelManagerData::reset()
• JDC record_upstream_sent_bytes() eliminates double-lock on sends

SV1 byte tracking (translator only):

• Arc counters in ConnectionSV1, shared with DownstreamData
• Reader counts line.len()+1 (newline stripped by LinesCodec)
• Writer counts serialized JSON bytes before write_all

Prometheus metrics (fixed cardinality):

• sv2_server_bytes_{received,sent}_total — aggregate scalar Gauges
• sv2_client_bytes_{received,sent}_total — aggregate scalar Gauges
• sv1_client_bytes_{received,sent}_total — per-client GaugeVec
  (1:1 with TCP connections, appropriate granularity)

Per-channel byte detail remains available via the JSON REST API
(/api/v1/server/channels, /api/v1/clients/{id}/channels,
/api/v1/sv1/clients) for drill-down without inflating Prometheus
time series cardinality.

Aggregate byte totals added to ServerSummary and Sv2ClientsSummary
so /api/v1/global includes byte totals alongside hashrate and
channel counts.