docs(fuse): add per-instance IO and process resource panels (#897)

[yangcx000]

Summary

Three additive enhancements to the curvine-fuse Grafana dashboard shipped in #1004: per-instance IO panels, a process-resources row, and Time-column de-duplication in tables. No existing panels are modified.

Issue Describe / Design

Related to #897 (end-to-end metrics; this is the fuse dashboard's consumption side). Follow-up to #1004.

Design notes:

• Per-instance IO uses sum by(instance,io_type) so an $instance=All selection no longer collapses read/write into one fleet line — each instance's throughput (B/s, success bytes) and IOPS (io_requests/s) is visible. Existing fleet-total panels are kept.
• Process resources reads the Prometheus client library's process_* metrics (from /proc, present even when metrics_enabled=false), filtered to job=~"$job" (fuse only) to stay within the dashboard's scope boundary. CPU is in cores, memory shows RSS vs virtual, plus open-fds-vs-max and OS threads.
• Time columns: multi-query tables carry one Time field per query; the merge-based read/write comparison table rendered Time 1..4. Hidden via an organize excludeByName transform on all five tables (existing column renames preserved).

Changes

Module / File	Change	Impact on existing behavior
etc/grafana/curvine-fuse-dashboard.json	+7 panels (ids 80–86): per-instance throughput/IOPS in the Stream IO row, new "Process resources (per-instance)" row (CPU/RSS/fds/threads); excludeByName Time transform on 5 tables	None — additive; existing panels unchanged
etc/grafana/README.md	Document per-instance vs fleet IO, the process-resources row and its scope	None

Test verified

Test case	Result	Notes
jq empty on dashboard JSON	PASS	valid; panel ids unique; only ids 80–86 added, none removed
Import to Grafana v12.2.1 + Prometheus	PASS	version bumped, all new panels present
Per-instance IO queries via datasource proxy	PASS	per-instance read/write throughput + IOPS resolve across 6 fuse instances
Process resource queries via datasource proxy	PASS	CPU (cores), RSS, open fds, threads all return per-instance data

Dependencies

• Related PRs: docs(fuse): add Grafana dashboard for fuse metrics #1004 (initial dashboard)
• Related issues: [FEATURE] End-to-end metrics across curvine-fuse, master, and worker #897