BF: always save data in <f4

[MarcCote]

Fixes #543
It seems that index trick np.r_[...] (for arrays concatenation) results in an array that has the same endianness of the system without any regard to the endianness of the arrays to be concatenated.

For instance, on a little-endian system:

In [1]: import numpy as np
In [2]: A = np.array([1,2,3], dtype=">f4")
In [3]: B = np.array([1,2,3], dtype=">f4")
In [4]: np.r_[A, B].dtype
Out[4]: dtype('float32')
In [5]: np.r_[A, B].dtype.byteorder
Out[5]: '='

[matthew-brett]

Save a line with:

data = np.r_[t.streamline, self.FIBER_DELIMITER]

?

[coveralls]

Coverage remained the same at 96.227% when pulling 25d5ba0 on MarcCote:bf_543 into c6e06d9 on nipy:master.

[codecov-io]

Codecov Report

❗ No coverage uploaded for pull request base (master@c6e06d9). Click here to learn what that means.
The diff coverage is 100%.

@@            Coverage Diff            @@
##             master     #545   +/-   ##
=========================================
  Coverage          ?   94.27%           
=========================================
  Files             ?      177           
  Lines             ?    24332           
  Branches          ?     2610           
=========================================
  Hits              ?    22940           
  Misses            ?      918           
  Partials          ?      474

Impacted Files	Coverage Δ
nibabel/streamlines/tck.py	98.88% <100%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update c6e06d9...c8b3479. Read the comment docs.

[coveralls]

Coverage decreased (-0.0002%) to 96.227% when pulling c8b3479 on MarcCote:bf_543 into c6e06d9 on nipy:master.

[matthew-brett]

Thanks for indulging my pedantry :)