scaling on partial model fits. Also added new method
isFirstLevelAnalysis.

design/stats.

ordered dict.

FeatImage.fit (caused I think by a numpy bug).

directory given a path, and to get top level analysis directory.

hopefully less buggy.

2. Updated documentation references to __init__ methods, as sphinx
   doesn't resolve them.

3. Bug fix in featresults module when loading contrast matrix.

appropriately. Some things are broken due to weakref-related changes in
the props package.

.feat/filtered_func_dat is loaded. This was the previous desired
behaviour, but there was a bug in it.

for cluster masks (and other random label images).

add Z statistics and cluster masks as overlays.

clustering has been run or not, as design.fsf settings are very
fickle. Instead, I am using the existence of cluster_*txt to determine
whether or not they exist.

DisplayContext.getDisplay was potentially creating Display/Opts
instances for overlays that were not in overlay list. This has been
rectified.

now in a module called featresults.py, which allows query/load of
settings, files, etc.

Also added ability to load an overlay from a FEAT directory (instead of
having to specify the filtered_func_data). But WX does not have a file
dialog which allows selection of either files or directories, so this
new feature can only be done via command line at present.

demeaning, or scaling to percent-changed. Added ability to plot data
reduced against a PE/COPE, and the ability to hide the raw data plot for
a currently-selected FEAT time series (as the plot can get pretty messy).

and residuals.

sum. 'Current time course' settings only shown if current time course is
being shown. Other little changes.

design a bit. Partial model fits are not being calculated correctly.

stuff. Overlay load function auto-detects FEAT-like images and creates
FEATImage instances instead of Image instances where necessary.