Token Information and Price Service

Problem/Feature Description

You are building a token research and portfolio tracking tool for Solana. The service needs two main capabilities:

1. Token Lookup: Users can search for tokens by name, symbol, or mint address. The UI displays token details, and the system needs to surface trust and safety information to help users avoid suspicious tokens. The system should treat some token identifiers as more authoritative than others.

2. Price Service: The tool fetches current prices for lists of tokens and uses them for portfolio valuation. Some prices are used in safety-critical contexts (e.g., deciding whether to trigger a stop-loss), so the system must handle cases where price data is unreliable or missing. The service needs to handle large batches of tokens efficiently.

The tool should be defensive: not all tokens have reliable data, some tokens may be flagged as suspicious, and the price API may return incomplete results for some tokens. The system should handle all of these gracefully rather than crashing or showing misleading data.

Output Specification

Produce a single TypeScript file called token-price-service.ts that exports:

1. searchTokens(query: string) - Search for tokens by name, symbol, or mint address
2. getTokenInfo(mintAddress: string) - Get detailed information for a single token
3. getTokensByTag(tag: string) - Get tokens by tag (e.g., verified, lst)
4. getTrending(category: string, interval: string) - Get trending/top tokens
5. getPrices(mintAddresses: string[]) - Get current prices for a list of tokens, handling batching if needed
6. getSafePrice(mintAddress: string) - Get a price with confidence validation suitable for safety-critical decisions
7. A TokenInfo interface that represents the normalized token data the UI will consume
8. A PriceResult interface that represents price data with confidence metadata

The code should use the native fetch API. It does not need to compile or run -- focus on correct API patterns, proper handling of missing/unreliable data, and defensive coding for the safety-critical price path.