Tracelines with Xsections

[dbrakenhoff]

From the example in #124. If I try to trace lines (after the adjustments in #125) backwards from the well, I'm getting reasonable results:

However, forwards tracking isn't working. The particles in the right Xsection are getting caught on the boundary and the (near) zero-thickness leaky layer.

• At the boundary the computation might be getting caught on the boundary between the inhoms, at the constant head/flux elements? But I don't understand why backward tracking isn't affected...
• Adjusting the vstepfrac should help with the very thin/zero-thickness aquitard, but setting this value to >1e10 didn't help. Making the aquitard at least 1E-04 m thick did help. This is probably something we should document somewhere.

So a few questions:

1. Which elements require changetrace() methods? At minimum I think 1D linesinks need one.
2. The XsectionsAreaSink's had old copied changetrace methods from the CircAreaSink class, which I modified in Refactor TimML plotting #125. I don't think the XsectionAreaSinkInhom needs a changetrace, since the boundary elements could theoretically take care of the changetraces? But the XsectionAreaSink might?
3. Can we get away with only setting changetrace on 1D LineSinks or do we need some special solution for Xsections?