[Analytics] Add platform-level latency metrics endpoint

[PierreLeGuen]

Description

Add a new endpoint to expose platform-wide latency metrics for the admin dashboard.

Motivation

Currently, latency metrics (TTFT, ITL) are only available at the organization level. Having platform-wide aggregates would help admins monitor overall system performance.

Proposed Solution

Add GET /v1/admin/platform/latency endpoint that returns:

• Average TTFT across all requests
• P95/P99 TTFT
• Average ITL (Inter-Token Latency)
• P95/P99 ITL
• Breakdown by model

Technical Details

• Query the organization_usage_log table for latency aggregates
• Support time range filtering (start/end params)
• Consider caching for performance

Acceptance Criteria

• New endpoint returns platform-wide latency metrics
• Supports time range filtering
• Returns breakdown by model
• Documentation updated