fix: re-enable legacy metrics reporter for audit bootstrap

[mateeullahmalik]

Problem

The audit module's epoch-end recovery for POSTPONED supernodes requires peer observations from active probers. On testnet, a deadlock exists where:

1. All supernodes running v2.4.5-testnet were already POSTPONED before the chain upgrade (legacy staleness kicked them ~500 blocks after they stopped submitting MsgReportSupernodeMetrics)
2. The epoch anchor's active_supernode_accounts has been empty since epoch 1 — no active probers exist
3. POSTPONED SNs faithfully submit audit epoch reports (~100 each) but cannot recover without peer observations
4. The 3 SNs on old releases (v2.4.29/35/36) bounce ACTIVE↔POSTPONED via legacy metrics every epoch but are always POSTPONED at epoch-start anchor freeze time

Fix

Re-enable the legacy supernode_metrics reporter alongside the audit host_reporter. Both run in parallel:

• Legacy MsgReportSupernodeMetrics → instant recovery from POSTPONED → ACTIVE (mid-epoch)
• Audit MsgSubmitEpochReport → prevents re-postponement at epoch end (report exists, host minimums disabled)
• Next epoch: SN is ACTIVE at anchor freeze → enters active_supernode_accounts → becomes a prober
• Peer observations flow → remaining POSTPONED SNs recover via audit path

Evidence (testnet lumera-testnet-2)

• 47/59 supernodes POSTPONED before audit activation (h=3763000)
• 12 SNs postponed at h≈3730xxx — exactly 501 blocks after last legacy metrics (staleness threshold = 500)
• 0 epoch reports from 3 old-release SNs (no audit code) — 229 bounce cycles each
• ~100 epoch reports from each v2.4.5-testnet SN — all stuck POSTPONED, zero peer observations available
• 0 active_supernode_accounts in every epoch anchor since epoch 1

After stabilization

Once the active set is stable (1-2 epochs), the legacy reporter can be removed in a future release.

[roomote-v0]

Rooviewer   See task

Reviewed the diff and traced the Collector type, its Run/Stop lifecycle, the service interface contract, the reachability.SetCurrentEpochID concurrency model, and the host_reporter interaction. No issues found -- the change is a clean re-enablement of previously working code with correct service interface compliance and safe concurrent behavior.

_{Mention @roomote in a comment to request specific changes to this pull request or fix all unresolved issues.}