Slow.

[rainwoodman]

Running the tinypy compiler under tpvm is very slow -- suggesting tpvm is much slower than cpython.

This needs some serious investigation.
The current evidence is that

• when I run tpy tests/test_args.py with tracing, I see tp->steps goes to up to a few million. The script takes a few seconds to finish.
• When I directly run the compiled bytecode with tpvm tests/test_args.tpc, it only takes less than a second. The vast majority of time is spent in running the compiler.

With memory errors gradually rooted out, we shall start some serious profiling.