Enable propagation backwards in imaginary time

--- old
+++ new
@@ -3,3 +3,5 @@
 In contrast to what is written in the documentation, propagating backwards in imaginary time is not unstable; it will generally converge to the largest eigenstate with a nrom growing to infinity, but that is a feature, not a bug.

 I am not quite sure what propagation in negative imaginary time is good for (increasing temperature? Spin states?), but it seems wrong to restrict the propagators that way for no good reason. While not really urgent, the cost is rather low.
+
+* Note: also update the documentation

--- old
+++ new
@@ -5,3 +5,7 @@
 I am not quite sure what propagation in negative imaginary time is good for (increasing temperature? Spin states?), but it seems wrong to restrict the propagators that way for no good reason. While not really urgent, the cost is rather low.

 * Note: also update the documentation
+
+----
+
+As an additional piece of work, removed OdeRelax. There is little point in this class, because relaxation with Chebychev polynomials should always be faster.