Bug: a88/02cc

Programmatically pick out the demagnetization level corresponding
to a one-half reduction in intensity, show it (optionally) on the
graph(s), and write it to a file for a whole suite.

Just closed #64 as dupe of this. Adding #64's description and 
comment here since it contains some information not included in this
ticket:

Christian writes: "A cunning trick to teach the puffin plot would
be a 'find mid point' on intensity plots.  This would be great
for NRM and ARM plots. There are scripts out there that will do
this for you. I have not yet put in the effort to learn how they
work but have grand plans."

I wrote: "Probably straightforward, but need to discuss with
Christian to get an exact definition for `midpoint', a rationale
for doing it, and references for the algorithm.- edited
description (pont, four hours ago)- created (pont, five hours
ago)"

Christian's keen on this and it's probably easy, so will try to
get it in for the 1.0 release.

Further details following discussion with Christian:

What we are looking to calculate is known as the `median destructive
field' (or, presumably, temperature, though this seems less common),
which is the field required to reduce the intensity to one half of its
initial value. It's described in Dunlop and Özdemir (p. 230). So all
we need to do is determine where the intensity curve crosses the
horizontal line corresponding to half the initial intensity, and read
off the corresponding field strength (or temperature). Interpolating
between two demagnetization levels will probably be necessary.

It's a little more complicated than that, though. Firstly, think about
NRMs with multiple components in different directions. The absolute
intensity might go up before it goes down, as the softest component
disappears first. What do we do then? There is no obvious right
answer. The way to implement it is probably to use the initial
intensity by default, but allow the user to manually recalculate using
only selected points.

Secondly, what if there's more than one crossing? Probably easy: pick
the first (likely to be the one we care about), and again have manual
restricted recalculation as a fall-back.

Thirdly, what if there's no crossing (i.e. we never get down to half
intensity)? Just go with the maximum field/temperature, says
Christian: it will stand out anyway when graphed. But perhaps it
should be marked in some other way when saving the values.

MDF should be marked on the graph by orthogonal lines projecting from
the axes (as is Tauxe lecture 9) and saveable as CSV (can presumably
be thrown in with PCA and Fisher values).

Now mostly done. After discussion with Christian, have decided on
initial implementation whereby the algorithm will be applied to 
all visible points. Works well in the simple case; less so if one
had multiple components with different directions. But we can
deal with that if and when it becomes necessary.

The other remaining task: label MDF value on X-axis of demagnetization
plot.

MDF now shown on X axis. Fully implemented, so closing.