Handle text-mode file objects in parse_wpc_surface_bulletin

[gaoflow]

Description

parse_wpc_surface_bulletin documents accepting a "str or file-like object", but it unconditionally calls .decode('utf-8') on the result of file.read(). open_as_needed opens filename paths in binary mode, but a file-like object passed in directly is returned untouched — so a text-mode object such as io.StringIO makes read() return str, and the .decode() call raises:

AttributeError: 'str' object has no attribute 'decode'

Fixes #3923.

Reproduction

from io import StringIO
from metpy.io import parse_wpc_surface_bulletin

sio = StringIO('''VALID 062818Z
HIGHS 1022 3961069 1020 3851069 1026 3750773
LOWS 1016 4510934 1002 3441145 1003 4271229
TROF 2971023 2831018 2691008
 ''')
parse_wpc_surface_bulletin(sio, year=2000)   # AttributeError before this PR

Fix

Only decode when read() returns bytes:

text = file.read()
if isinstance(text, bytes):
    text = text.decode('utf-8')

I went with an explicit isinstance(text, bytes) check rather than a broad try/except AttributeError (discussed in the issue): it handles exactly the two things the docstring promises — binary sources ('rb' file paths, BytesIO) and text sources (StringIO, text-mode files) — without swallowing unrelated AttributeErrors or revisiting the Python 2→3 byte-handling story more broadly.

Tests / docs

• Added test_parse_wpc_surface_bulletin_text_file_object, asserting a StringIO parses identically (pd.testing.assert_frame_equal) to the equivalent BytesIO.
• Moved the misplaced year entry from the Returns section to Parameters in the docstring (it is a parameter, not a return value).

tests/io/test_text.py (all 5), the text.py doctest, and ruff pass locally.