[client]: make batch request easiest to understand when a too large response occurs

[niklasad1]

After some usages of the batch request API and if one constructs a too big request or response it's not easy to understand.
For instance the jsonrpsee server will reply with ["batch too large", "id": null] which we doesn't handle differently in the clients.

For the async-client/websocket client one would get some error message such Invalid RequestId which isn't user-friendly and a better log/error should be propagated to the user.

For the http client is possible to detect by matching on the error but one really have to downcast the error which isn't good either.

Thus,

1. Provide concrete error types/variants instead of using anyhow::Error
2. Add a some concrete error variant InvalidBatchResponse

Optionally, add some documentation that batch requests might bump into the max_request, max_response limits.