Gracefully handle FileHistory decoding errors

[op3]

In case of a corrupted history file, instead of raising a UnicodeDecodeError, malformed characters are replaced by U+FFFD (REPLACEMENT CHARACTER) during decoding.

Currently, if a history file is malformed and contains invalid unicode characters, a UnicodeDecodeError is raised, which prevents the prompt from working (and cannot be handled properly).

Minimal working example (before fix):

Create malformed history

echo -n -e '\xe9' >> history

Example code:

#!/usr/bin/env python

from prompt_toolkit.shortcuts import prompt
from prompt_toolkit.history import FileHistory

prompt(history=FileHistory("history"))

Execution and output before fix:

$ ./mwe.py
Traceback (most recent call last):
  File "/tmp/./mwe.py", line 6, in <module>
    prompt(history=FileHistory("history"))
  File "/usr/lib/python3.9/site-packages/prompt_toolkit/shortcuts/prompt.py", line 1382, in prompt
    session: PromptSession[str] = PromptSession(history=history)
  File "/usr/lib/python3.9/site-packages/prompt_toolkit/shortcuts/prompt.py", line 463, in __init__
    self.default_buffer = self._create_default_buffer()
  File "/usr/lib/python3.9/site-packages/prompt_toolkit/shortcuts/prompt.py", line 498, in _create_default_buffer
    return Buffer(
  File "/usr/lib/python3.9/site-packages/prompt_toolkit/buffer.py", line 316, in __init__
    self.history.load(new_history_item)
  File "/usr/lib/python3.9/site-packages/prompt_toolkit/history.py", line 70, in load
    for item in self.load_history_strings():
  File "/usr/lib/python3.9/site-packages/prompt_toolkit/history.py", line 210, in load_history_strings
    line = line_bytes.decode("utf-8")
UnicodeDecodeError: 'utf-8' codec can't decode byte 0xe9 in position 0: unexpected end of data

[jonathanslenders]

Thanks!