Change to ParaParser: well formed tags now required

Hi everyone,

In the course of adding in our dormant RTL (Arabic support) branch, we
had to fix a weirdness in the paragraph parser.  Henceforth we will
require well formed XML tags.

For many years we have silently allowed people to get away with things like

   Hello <b>malformed</i> world

We can't do this any more while accurately tracking the state of the
text, so henceforth we will raise a ValueError.

It doesn't break any tests, since we never thought to assert that we
should accept such content.   It won't affect any commercial customers
using RML, as the input is always strictly validated.   However, if
you do have an application which might be bitten by this, please do
let us know.    Some kind of preprocessing or cleanup utility could
potentially be provided or used.

Best Regards,


-- 
Andy Robinson
Managing Director
ReportLab Europe Ltd.