unicode lazy strings don't work on Jython

[leosoto <leo.soto@…>]

The problem is that on Jython __unicode__ lives on object (just like __str__ ), so ​assigning the __unicode__ attribute to lazy instances don't work there.

[leosoto <leo.soto@…>]

Simple patch, works ok on CPython.

[leosoto <leo.soto@…>]

Well, Jython now dropped object.__unicode__, but a __unicode__ method assigned specifically to an instance doesn't work.

In fact, on CPython it seems that the only __*__ method that works when assigned to an instance is __unicode__ (all the others are looked up on the type, not the instance).

So, this patch still sounds reasonable, as the original code relies on a obscure implementation detail of CPython.

[Malcolm Tredinnick]

Nice idea. I hadn't realised the significance of creating a new __proxy__ each time. This is much neater than what we're doing in __str__, so I'll change that, too.

[Malcolm Tredinnick]

(In [6587]) Fixed #5560 -- Improved the way we create str and unicode methods in
lazy() objects. This fixes things for Jython and makes the code more readable,
even for CPython. Thanks, Leo Soto.