fix: unescape escape sequences in double-quoted section header keys

[gaoflow]

Bug

TOML spec §2.4 (Keys) says quoted keys follow
the same escaping rules as basic strings. The section-header parsing loop in
decoder.py stripped surrounding quotes from double-quoted key names but never
called _unescape(). As a result escape sequences like \t, \n, \r, and
\\ were left as literal backslash-letter pairs.

import toml, tomllib   # tomllib is the stdlib reference implementation

toml_str = '["a\\tb"]'  # key should contain a tab character

key_toml    = list(toml.loads(toml_str).keys())[0]
key_tomllib = list(tomllib.loads(toml_str).keys())[0]

print(len(key_toml))    # 4  ("a" + "\" + "t" + "b") — wrong
print(len(key_tomllib)) # 3  ("a" + TAB  + "b")       — correct

The same _unescape() call that load_line already makes for inline
double-quoted keys (line 769) was simply missing for section-header keys.

Fix

In toml/decoder.py, the quoted-key stripping inside the group-name loop
now calls _unescape() for double-quoted key names. Single-quoted keys
are left unchanged (TOML spec: literal, no escaping).

Tests

Added test_section_header_key_escape_sequences which checks \t, \n,
\r, and \\ against the stdlib tomllib reference. All 25 tests pass.

This pull request was prepared with the assistance of AI, under my direction and review.