Default ranges in hough_line now have integer stepsizes

[hohlraum]

Addresses issue #1433.

Default theta runs in the range [-pi/2, pi/2) in 1deg steps.
Default distances runs in the range [-diag, +diag) in 1 px steps, where diag is the length of the image diagonal.

[JDWarner]

It appears the previous behavior, which landed between integer values, was hard coded into two tests. Travis is failing because of these:

======================================================================
FAIL: skimage.transform.tests.test_hough_transform.test_hough_line
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/travis/venv/local/lib/python2.7/site-packages/nose/case.py", line 197, in runTest
    self.test(*self.arg)
  File "/home/travis/build/scikit-image/scikit-image/skimage/transform/tests/test_hough_transform.py", line 30, in test_hough_line
    assert_almost_equal(dist, 80.723, 1)
  File "/usr/lib/python2.7/dist-packages/numpy/testing/utils.py", line 468, in assert_almost_equal
    raise AssertionError(msg)
AssertionError: 
Arrays are not almost equal to 1 decimals
 ACTUAL: 80.0
 DESIRED: 80.723
======================================================================
FAIL: skimage.transform.tests.test_hough_transform.test_hough_line_peaks
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/travis/venv/local/lib/python2.7/site-packages/nose/case.py", line 197, in runTest
    self.test(*self.arg)
  File "/home/travis/build/scikit-image/scikit-image/skimage/transform/tests/test_hough_transform.py", line 75, in test_hough_line_peaks
    assert_almost_equal(dist[0], 80.723, 1)
  File "/usr/lib/python2.7/dist-packages/numpy/testing/utils.py", line 468, in assert_almost_equal
    raise AssertionError(msg)
AssertionError: 
Arrays are not almost equal to 1 decimals
 ACTUAL: 81.0
 DESIRED: 80.723

I'm in favor of modifying the tests. Anyone else from @scikit-image/core want to chime in?

[blink1073]

👍 for updating what amounts to a regression test.

[bennlich]

It seems strange that the two tests don't produce the same number. Both are run on the same image: compare L21-L24 with L67-L69.