…lues.

Ensure that if the connectome matrix selection is updated, the limits for relevant controls are also updated.
Also functionalised some redundant code to compress things a little.

The original version was based on the duality principle for cases m<n and m>n, which in some cases led to asymmetric numerical instabilities. Here, we modify the formulas to be consistent with the m>n case at all times, which we found to be most numerically stable.

The formulas are now always made consistent with the m<n case that is discussed in tha paper, effectively taking the transpose whenever m>n.

We chose to keep the patch size identical, hence using a block 'against'
the edge of the image to estimate the noise level and then denoising the
non-centred voxels.

…onding .cpp file found #1042

…t nitpickery

… for speed reasons. Files without shebang in first line or non UTF-8 encoding in first 1024 bytes get deleted on build.

… of deleting them

See #1604.

Fixes for TravisCI on dev

Revert changes that were made in #1507 with respect to determination of the default maximum angle for the various tracking algorithms. There, this decision was changed from being based on (90deg * step_size / vox()) to being a minimum radius of curvature of one voxel. Additionally, the default step size in the presence of RK4 was changed from 0.1*vox() to 0.5*vox(). This commit reverts these changes, but retains the calculation of minimum radius of curvature necessary for streamline length calculation optimisations.

tcken: Revert default maximum angle calculation

docs: fixed SH description URL

Prevents implicit construction of Header based on input.

Conflicts:
	lib/mrtrix3/image.py
	lib/mrtrix3/path.py

JSON handling changes

Python: Use of kwargs for API functions

Build remove outdated bin

Mtnormalise tests

dwifslpreproc: Add help description RE: CUDA