D4: Top processes by CPU / memory

[LarsLaskowski]

Epic: More Metrics

Why

When CPU or RAM spikes, users want to know what is responsible.

Scope

Walk /proc/[pid]/stat and /proc/[pid]/status (or comm), compute per-process CPU% via deltas, and return the top-N by CPU and by RSS. Bound N and compute on the slow tick to limit cost.

Reuse & touch points

• New internal/collector/process.go + fixtures (feed synthetic /proc/<pid> directories via testhelpers_test.go).

API & config impact

• New GET /api/v1/processes (kept off the main snapshot to bound payload size).
• Config processes_enabled, processes_top_n.

Acceptance criteria

• A fixture pid set produces the correct top-N ordering by CPU and by memory.