feat: implement Stellar network fee estimation caching

[Phantomcall]

Stellar Network Fee Estimation Caching

Problem

Every payment initiation currently fetches the current fee from Stellar Horizon synchronously via server.fetchBaseFee(), adding latency to each transaction. This is particularly impactful for high-throughput payment processing where every millisecond counts.

Solution

Implement a Redis-backed caching layer for Stellar Horizon fee statistics with a 30-second TTL, refreshed by a background cron job. The payment service reads from the cache instead of making a Horizon call per payment, eliminating redundant network round-trips.

Changes

New Files

File	Purpose
src/services/stellarFeeStatsCache.ts	Cache service: reads fee stats from Redis (with Horizon fallback on cache miss), fetches fresh stats from Horizon, writes to Redis with 30s TTL
src/jobs/stellarFeeStatsJob.ts	Background cron job that refreshes the fee stats cache every 30 seconds

Modified Files

File	Change
src/stellar/transactions.ts	getTransactionBaseFee() now reads from Redis cache first; falls back to server.fetchBaseFee() on cache miss
src/jobs/scheduler.ts	Registered the new stellar-fee-stats job at */30 * * * * *
src/index.ts	Added fee stats cache warm-up on startup (after Redis connects)

How It Works

1. Startup: On application start, after Redis connects, the fee stats cache is immediately populated by fetching from Horizon and storing in Redis.
2. Background refresh: A cron job runs every 30 seconds, fetches server.feeStats() from Horizon, and writes last_ledger_base_fee to Redis under key stellar:fee_stats with a 30-second TTL.
3. Payment flow: When building a transaction, getTransactionBaseFee() attempts to read from Redis first. If the key exists, it returns the cached value immediately (no network call). On cache miss, it falls back to the original server.fetchBaseFee().
4. Logging: Cache values are logged at startup and on each background refresh for observability.

Acceptance Criteria

• Background job fetches Horizon fee stats every 30 seconds and stores in Redis
• Payment service reads fee from Redis cache instead of calling Horizon per request
• On cache miss, fee is fetched directly from Horizon as a fallback
• Cached fee values are logged at startup and on each refresh

closes #113

[drips-wave]

@Phantomcall Great news! 🎉 Based on an automated assessment of this PR, the linked Wave issue(s) no longer count against your application limits.

You can now already apply to more issues while waiting for a review of this PR. Keep up the great work! 🚀

Learn more about application limits