ObjectId creation from BSON could be optimized

When decoding a BSON string the generation of ObjectIds could be optimized to avoid several function calls.

For example, instead of creating ObjectIds the BSON decoder could create objects of this class:

    class _ObjectIdFromString(ObjectId):

        def __init__(self, oid):

            self._ObjectId__id = oid

This new class is just providing a simplification of ObjectId.__init__ for the case when the input is known to be a sequence of 12 bytes.

The changes to the c extension would be trivial, the attached patch file shows the 4 lines that require changes.