documentation and fix trace for cvrisk of noncyclical fitting

[ja-thomas]

This should fix #16

We cannot print the trace in the same way as in the cyclical algorithm. Since we do not update the model step by step but rather fit the largest mstop and than extract the submodels.

I could change that behavior, but I think this will be a bit more expensive in terms of calculation time, which is a bad trade of for a nicer printer I would say.

[coveralls]

Coverage increased (+0.03%) to 79.03% when pulling d62ff7d3e2401d0b6a4fc3eac38fa0ab9bcc03df on fix_16_noncyc into e4b7cd07060fa34f9fdecd415565f012b2794f47 on devel.

[hofnerb]

• Can't we simply set trace = trace in the update(mod) step? Additionally, we could print bevore the update one line that states the current cross-validation iteration. Wouldn't that be nicer? At least it currently makes no sense:

Starting cross-validation...
[fold]  [current mstop]
 [1]    [62,38]
 [2]    [53,47]
 [3]    [55,45]
 [4]    [58,42]
 [5]    [67,33]
 [6]    [65,35]
 [7]    [66,34]
...

• Additionally, the result of mstop() is buggy. See #19

[ja-thomas]

Starting cross-validation...
[fold: 1]
[   1]  -- risk: 1237.839 
[  4] .......................................................... -- risk: 1186.652 
[ 64] ..................................
Final risk: 1169.155 
.
[fold: 2]
[   1]  -- risk: 1275.722 
[  4] .......................................................... -- risk: 1228.878 
[ 64] ..................................
Final risk: 1212.287 
.
[fold: 3]
[   1]  -- risk: 1208.189 
[  4] .......................................................... -- risk: 1155.631 
[ 64] ..................................
Final risk: 1140.451 
.
[fold: 4]
[   1]  -- risk: 1266.236 
[  4] .......................................................... -- risk: 1195.522 
[ 64] ..................................
Final risk: 1179.904 
.
[fold: 5]
[   1]  -- risk: 1179.009 
[  4] .......................................................... -- risk: 1126.048 
[ 64] ..................................
Final risk: 1110.435 
.

There are still two formatting errors

the bracket for [ 1] is to large and there is a strange additional point between every fold, still have to fix that.

[coveralls]

Coverage increased (+2.2%) to 81.201% when pulling 674e677fd7f8efc390b3a373f96a9e467de4b371 on fix_16_noncyc into e4b7cd07060fa34f9fdecd415565f012b2794f47 on devel.

[ja-thomas]

@hofnerb Should be good to go now