sharded.hh: add invoke_on variant for a shard range

[tomershafir]

Add a convenient method to invoke a function on a range of shards, to have better support for the cold control path on more complex compute models like shard groups for different tasks. A workload can benefit from such a model if inter-task cooperation is better when grouping from perf and QoS perspectives, for example. Or a computation requires a only a subset of shards due to internal concurrency limits.

[tomershafir]

@avikivity changed to std::ranges::range