tomllib reports an error position beyond EOF for some files ending in backslash #158566

Bug description

This is a fairly minor bug in tomllib: when a document ends with an unfinished backslash escape inside a double-quoted string or key, the exception reports that the error is past the end of the document.

import tomllib

document = '"\\'  # A quote followed by a backslash.
try:
    tomllib.loads(document)
except tomllib.TOMLDecodeError as exc:
    print(len(exc.doc), exc.pos, exc.lineno, exc.colno)
    # 2 3 1 4
    # Doc length = 2
    # Error position = 3 (1 past the EOF)
    # Line number = 1 (correct)
    # Column number = 4 (1 past EOF)

The document has two characters, but pos is 3. I would expect the position to point either to the backslash or to EOF, which would be index 2 and column 3. As a practical matter the formatted message already says "at end of document", so this only affects callers using the location attributes.

The same thing happens for a trailing backslash in a basic string value or a multiline basic string. An unterminated string without the backslash, such as '"abc', reports pos == len(doc) as expected.

The cause is parse_basic_str_escape advancing the position by two before checking the escape, even when only the backslash remains. That position is then used to construct TOMLDecodeError.

I found this while adding Hypothesis tests in gh-158565. The PR includes a focused expected-failure property and the minimized example above. This is related to but distinct from Tomli #307, which concerns the location reported for invalid Unicode escape digits.

CPython versions tested on

• Python 3.14.7.
• Main at eb30e3d9d4b523e3f8de2cd1485c9ae9a996ab03, using a local Python 3.16.0a0 build to run that checkout's tomllib.

Operating systems tested on

Linux.