Do not try to put to blobstor on full write cache

Is your feature request related to a problem? Please describe.

I'm always frustrated when I'm thinking of write cache and blobstor relations. Currently we always try to put object to blobstor if putting to write cache fails.

Describe the solution you'd like

Don't do that in case write cache is full. If that's the case it's supposed to be flushing to blobstor at the maximum speed possible and additional threads pushing objects to blobstor directly can only slow it down. If it's full because it can't flush objects to blobstor trying this direct put doesn't help either.

Describe alternatives you've considered

Keep things as is. Current behavior is somewhat better for the case when write cache can't put objects to blobstor, but direct put somehow can. I don't see this happening, but who knows.

Additional context

https://github.com/nspcc-dev/neofs-node/pull/3744#discussion_r2635320251

No file or test path is provided; start by locating the write-cache put handling and its fallback to blobstor. Trace how a full cache is distinguished from other write failures, then verify that a full cache no longer triggers a direct blobstor put while other fallback behavior remains unchanged.