Avoid erroring from tracing

A user reported a weird condition in which pluggy tracing was making their code fail after we enabled tracing because one of the local objects didn't have a working __repr__ (a scikit-learn object from a very old version). Details: https://github.com/kedro-org/kedro/issues/2630

We gave the user a workaround and also it's reasonable to expect that objects have a working __repr__, but it would be nice if there was a generous try ... except around the tracing messages so that errors there don't bubble up. Essentially the same that happens with logging:

>>> try:
...   logging.warning("Will fail: %d", "a")
...   logging.warning("But the show must go on")
... except:
...   logging.error("Oh no")
... else:
...   logging.warning("Everything is fine")
... 
--- Logging error ---
Traceback (most recent call last):
  File "/opt/homebrew/Cellar/python@3.11/3.11.4/Frameworks/Python.framework/Versions/3.11/lib/python3.11/logging/__init__.py", line 1110, in emit
    msg = self.format(record)
          ^^^^^^^^^^^^^^^^^^^
  File "/opt/homebrew/Cellar/python@3.11/3.11.4/Frameworks/Python.framework/Versions/3.11/lib/python3.11/logging/__init__.py", line 953, in format
    return fmt.format(record)
           ^^^^^^^^^^^^^^^^^^
  File "/opt/homebrew/Cellar/python@3.11/3.11.4/Frameworks/Python.framework/Versions/3.11/lib/python3.11/logging/__init__.py", line 687, in format
    record.message = record.getMessage()
                     ^^^^^^^^^^^^^^^^^^^
  File "/opt/homebrew/Cellar/python@3.11/3.11.4/Frameworks/Python.framework/Versions/3.11/lib/python3.11/logging/__init__.py", line 377, in getMessage
    msg = msg % self.args
          ~~~~^~~~~~~~~~~
TypeError: %d format: a real number is required, not str
Call stack:
  File "<stdin>", line 2, in <module>
Message: 'Will fail: %d'
Arguments: ('a',)
WARNING:root:But the show must go on
WARNING:root:Everything is fine

Start by locating pluggy's tracing implementation and the code that formats tracing messages; the issue does not name a file or test. Reproduce the failure with an object whose repr raises, then verify that tracing errors no longer propagate and normal application execution continues.