Provide an API that explains the query execution plan

[aleks-p]

It can be beneficial to have feature similar to SQL EXPLAIN, where a user, developer or an administrator can get an idea of how a particular query is executed. It could provide information such as

• the components involved (number of queriers, ingesters, store-gateways)
• the number of blocks that will be accessed by each component
• the (total) size of the blocks, number of series and profiles
• whether deduplication will happen or not

Usage:

• Can be called by a frontend before running a query to give the user an idea of how the query would be executed (similar to how Loki shows how much data would be processed by a query)
• Can be used to warn the users for potentially slow queries
• Can be used to provide suggestions on how to improve a query (e.g., based on Create config setting to customize an environment for querying #2796, or changing max nodes).
• Could be visualized in a node graph

Notes / considerations:

• Today this kind of analysis is done using tracing, which can be challenging for large traces