Further simplify DataColumnSidecarsByRoot request

[g11tech]

PR - #4284 simplified DataColumnSidecarsByRoot request.

this PR further proposes to even pull out columns from the identifiers because a client generally would request same missing columns for a set of roots and hence don't need to keep repeating the same columns again and again for reach block root.

this further shortens the size of request making p2p traffic lighter and should be an easy update to the api usage in the clients.
it may also help the responders to quickly analyze the columns that they need to respond to the request without iterating the entire request which my offer a small space for optimization as well.

[jtraglia]

Thanks for the PR! I think this change makes sense.

PS: please run make lint to fix the failing CI check.

[ppopth]

This PR makes sense to me. Having DataColumnsByRootIdentifier is quite redundant.

[tersec]

this further shortens the size of request making p2p traffic lighter

it may also help the responders to quickly analyze the columns that they need to respond to the request without iterating the entire request which my offer a small space for optimization as well.

can you quantify these? If it's an optimization, it should be quantitative.

[pawanjay176]

In Lighthouse, we only request a single root in all our byroot request. So this change isn't really useful for us imo.
Not sure if we are thinking of asking for multiple roots in the same request.
cc @dapplion

[jtraglia]

Hey @g11tech, it appears that this change doesn't have a lot of support.

[g11tech]

closing the PR as the current/future usage patterns don't warrant this optimization