src/crimson/pg_backend: truncate to 0 when EC pool in _writefull

This commit adds the missing EC branch in PGBackend::_writefull,
which truncates the existing object to 0 (ie deletes it) before
writing the new object.

Before this change, it only implemented the replicated pools
logic, ie to truncate the existing object to the length of the new
object. This logic, for EC, meant writefull for smaller new objects
is planned as an in-place read-modify-write. Now, parity needs to
be recomputed and we need clones to save the old data for rollbacks.
In seastore, this clone read hangs, leading to slow ops.

For EC, we should not be doing this for a writefull where a new
object needs to replace the old object.

Fix: simply truncate the old object (delete) before writing the new
object.

Fixes: https://tracker.ceph.com/issues/79453
Signed-off-by: Shraddha Agrawal <shraddha.agrawal000@gmail.com>