SFR2: Allow numpy to format floats

[mwtoews]

This fix is related to #300, where flopy's mfsfr2.write_file method adds superfluous digits to output files.

The issue seems to be with how Python's str.format interacts with Numpy. For instance, with two examples:

import numpy as np
x = np.float32(101.1)
print('{}'.format(x))  # 101.0999984741211
print('{!s}'.format(x))  # 101.1

With the first print example, numpy's float32 is converted to Python's native float, which is a double precision value that is formatted by Python as a native float. It is equivalent to print('{}'.format(float(x))) with lousy output. This is the same behaviour with Python 2 and 3.

With the second print example, numpy returns a string representation of the float32 type, retaining the correct precision. This PR implements the second method to let numpy format the value to a str. The example described in #300 is formatted nicer as:

# SFR package for  MODFLOW-2005, generated by Flopy.
4 2 0 0 86400.00000000 0.00010000 0 0 
1 1 1 0 0 101.1
1 1 1 0 0 202.2
1 1 1 0 0 103.4
1 1 1 0 0 193.2
2 0 0
0 0 0 0 543.4 0 0 0 
0 0 0 0 0 
0 0 0 0 0 
0 0 0 0 987.6 0 0 0 
0 0 0 0 0 
0 0 0 0 0 

Is there any performance difference? Not really:

$ python3
Python 3.5.3 (default, Jan 19 2017, 14:11:04) 
[GCC 6.3.0 20170118] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import timeit
>>> import numpy as np
>>> timeit.timeit("'{}'.format(x)", "import numpy as np; x = np.float32(101.1)", number=100000)
0.1474021040048683
>>> timeit.timeit("'{!s}'.format(x)", "import numpy as np; x = np.float32(101.1)", number=100000)
0.15334038599394262

[coveralls]

Coverage decreased (-1.8%) to 69.468% when pulling c3eb9e2 on mwtoews:sfr-mods2 into 05a9caf on modflowpy:develop.