`LocalSplitStore::open` can take a very long time when there is a high number of splits on the disk

For 100k splits, it took 1min to open... so what?

Well, the problem is the following: we start the REST server after LocalSplitStore::open... and if you are on Kubernetes, your pod will restart before it finishes the opening of the local split store.

Of course, we can tell Kubernetes to wait more than 1 minute before considering a restart but... it looks like a bad solution.

We can see several solutions to avoid this bad experience:

• first, we can add some logs to indicated how many splits there are and warn the user if it takes a long time
• secondly, we can set a max value to split_store_max_num_splits.
• thirdly, optimize the opening but I don't think it's necessary to do that now.

Start at the LocalSplitStore::open entry point and reproduce the reported one-minute startup with a large number of splits, such as 100k. Review the available approaches in the issue—startup logging, limiting split_store_max_num_splits, or optimizing opening—and clarify which outcome is required; done should prevent long startup from causing Kubernetes restarts.