feat: Use nullable Float64Dtype to allow NULL and NaN to be represented in the same Series when `dtype_backend="numpy_nullable"`

The Float64Dtype more closely matches the BigQuery semantics. Ideally we'd allow folks to use this to avoid the slight data loss of mapping both NULL and NaN to NaN with the existing numpy.float64 dtype.

Edit: Per googleapis/google-cloud-python#14482, we should only do this if dtype_backend="numpy_nullable".