HTTPS Outcall: Add another http_request method to improve the dependency on a single point of data

[iclighthouse]

The current http_request() method accepts a url parameter url (single point data), and all nodes in the subnet go to access this single point data and then form a consensus, this consensus only solves the problem of preventing data inconsistency during the access, and there is no consensus on the data itself. If I want to get the consensus of multi-point data, I need to access multiple urls (multi-point data) to re-execute the consensus at the business layer. But this brings the problem of inefficiency and excessive cycles consumption. Scenario: Canister accesses the RPC url of an ethereum node via HTTPS Outcall, and this time it runs the risk of single point data failure or false data. If I go to access the RPC of multiple nodes at the same time, this overhead and cycles consumption is very high. Is it possible to provide a http_request2() method where the url field in the parameters is changed to urls, an array of urls. Each node of the subnet randomly selects a url to access and then performs consensus on the data. This solves the above problem.

This is important for accessing distributed system that has the same API.

[dietersommer]

This is a good idea and I think we should keep this in mind for a future extension of the feature. Particularly, when using HTTPS outcalls to call to Ethereum RPC nodes, this may be of interest.