Feature request: numpy.random.normal with `dtype`

When constructing arrays it is useful to set the dtype. For some reason we can do it with some np.random methods such as random integers but not with random normal values. Could we have a consitent version across all methods that generate new np.arrays?

>>> np.random.randint(10,dtype="Int16")
9
>>> np.random.normal(10,dtype="Float32")
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "mtrand.pyx", line 1547, in mtrand.RandomState.normal (numpy/random/mtrand/mtrand.c:20493)
TypeError: normal() got an unexpected keyword argument 'dtype'

This would be nice in order to avoid np.array( somearray, dtype=sometype ) every time we create a new random array.

For example

np.array( np.random.normal(10), dtype="Float32")

could be simply written as

 np.random.normal(10, dtype="Float32")

[bashtage]

Use randomgen until the new PRNG infrastructure gets into NumPy (hopefully this year). 32 bit is supported for the most common types (normal, exponential, uniform and gamma).

[lucasb-eyer]

I would like to re-open the issue, as I keep bumping into wanting this almost weekly for over a year. The conversion is not a solution, because most of the times I need this it is for memory purposes. It seems to be a somewhat common request, see for example this SO question about it (not me), and the proposed workaround.