[SERVICES-2691] Performance improvements and caching fixes

[mad2sm0key]

Reasoning

• using falsy checks instead of explicit undefined checks inside the cache getAllKeys method triggers unnecessary individual redis GET requests for cached booleans / the number 0
• LP tokens metadata is only being cached on request resulting in slow queries for the first user that needs the data
• dataloaders have a default in-memory cache enabled

Proposed Changes

• disable caching for pair (abi + compute) and tokens dataloaders
• add dataloader for pair 24h volume
• cache warm LP tokens metadata
• replace falsy check with explicit undefined check in getAllKeys util function

How to test

• N/A