[Newcompiler] Finding enough bytecodes for closures.

[bryce at kampjes.demon.co.uk]

Klaus D. Witzel writes:
 > Bryce is an expert on contexts, I as you believe he meant the  
 > MethodContext which must exist before creation of some BlockContext makes  
 > any sense (any sense to the VM).
 > 
 > Nevertheless, I hope he clarifies :)

I did mean the MethodContext that created the BlockContext. There
is always the BlockContext and a surrounding context where the
block is initialised. (Though that surrounding context might be
another BlockContext, in which case it could't be recycled.)

I was also assuming that we were using an bytecode set at least as
efficient as Mathieu's current one that avoids directly accessing
thisContext.

Bryce