Use native Python double precision float for SFR2

[mwtoews]

Flopy's SFR2 package adds superfluous digits to output files from native Python float types, which is double precision. Let me demonstrate:

>>> import numpy as np
>>> import flopy
>>> m = flopy.modflow.Modflow()
>>> d = flopy.modflow.ModflowDis(m)
>>> s = flopy.modflow.ModflowSfr2(m, nstrm=4, nss=2)
>>> s.reach_data['rchlen'] = [101.1, 202.2, 103.4, 193.2]
>>> s.reach_data['rchlen']
array([ 101.09999847,  202.19999695,  103.40000153,  193.19999695], dtype=float32)
>>> s.segment_data[0]['flow'] = [543.4, 987.6]
>>> s.segment_data[0]['flow']
array([ 543.40002441,  987.59997559], dtype=float32)
>>> s.write_file('x.sfr')

And the x.sfr output file:

# SFR package for  MODFLOW-2005, generated by Flopy.
4 2 0 0 86400.00000000 0.00010000 0 0
1 1 1 0 0 101.099998474
1 1 1 0 0 202.199996948
1 1 1 0 0 103.400001526
1 1 1 0 0 193.199996948
2 0 0
0 0 0 0 543.400024414 0 0 0
0 0 0 0 0
0 0 0 0 0
0 0 0 0 987.599975586 0 0 0
0 0 0 0 0
0 0 0 0 0

The fix that I've put together in this PR is to switch from float32 (a.k.a. real) to float (a.k.a. double precision).

I haven't dug down to SFR2's writer/formatter, but it's possible to get Numpy to output float32 (or any other data type) as it should:

>>> for x in s.segment_data[0]['flow']:
...     print(x)
...
543.4
987.6
>>> print(s.segment_data[0]['flow'])
[ 543.40002441  987.59997559]

---

Other fixes in this PR add other floats to NC_PRECISION_TYPE. Also np.int64 was corrected from "i4" to "i8".

[coveralls]

Coverage remained the same at 69.807% when pulling 1a1e857 on mwtoews:sfr-mods into 0875d48 on modflowpy:develop.

[mwtoews]

Closing as this issue with SFR2 formatting is primarily addressed in #313

However, there are potentially outstanding issues with NC_PRECISION_TYPE, i.e. it should have np.int64: "i8"