[FIlestore] make entry_timeout configurable for regular files

[debnatkh]

A simple scenario that shows a problem with using non-zero entry_timeout:

1. client1 performs touch srcfile dstfile and caches two dentry records for each file: srcfile → inode1, dstfile → inode2.
2. client2 performs stat srcfile dstfile and caches two dentry records for each file: srcfile → inode1, dstfile → inode2.
3. client1 performs mv srcfile dstfile. inode2 is no longer present in the filesystem.
4. client2 performs cat dstfile. The guest kernel's dentry cache resolves the dstfile to inode2 and attempts
5. to open it. It fails since this inode is no longer present.

---

One possible workaround is to customize the entry_timeout setting for regular files, allowing us to lower it. It is reasonable to keep the default entry_timeout at 15 seconds, not to hinder effective dentry caching used for the resolution of /very/long/absolute/path/to/a/file.