Mutable writes don't respect `readonly` or `reserved_space` storage server settings

[exarkun]

There are simply no checks in the storage server implementation against these parts of the configuration. The storage server always attempts to satisfy mutable write attempts. They might fail because the filesystem is *actually* read-only or *actually* out of space but not because of the Tahoe-LAFS configuration for these things.

This means the expected end-state for any storage server is to completely fill the filesystem where its storage is located and then become non-operable because of the myriad ways in which software fails when encountering ENOSPC.