Tracing code traces the live-py code

Here's an example that shows how the user's code is drowned out by the live-py code. It never even gets to the `foo()` method before it hits the message limit.

```
import sys | RuntimeError: live coding message limit exceeded
|
|
def traceit(frame, event, arg): | event = 'call' arg = None | event = 'line' arg = None | event = 'line' arg = None | event = 'call' arg = None | event = 'line' arg = None | event = 'line' arg = None ...(trimmed)
location = frame.f_code.co_name, frame.f_lineno | location = ('record_call', 125) | location = ('record_call', 131) | location = ('record_call', 133) | location = ('record_call', 125) | location = ('record_call', 131) | location = ('record_call', 133) ...(trimmed)
return traceit | return | return | | return | return | return

Start by locating the live-py tracing setup and the code that handles the message limit, then inspect how sys.settrace is used. Reproduce the issue with the example in the report. Done means tracing the user's code reaches foo() without live-py frames drowning out the user's trace output.

Written by the indexing model from the issue text.