`pospell` fails with code entries

For 3.13, there was a change in the documentation that now snippets were marked as translatable.
This means that there is a large number of "new entries" that are snippets from the documentation page.

Even though I believe this is a great thing, the tooling around might need to be adapted.
Recently I discovered that in some cases, pospell fails to parse and handle such entries.

For example:

#: ../Doc/library/typing.rst:3167
msgid ""
"@type_check_only\n"
"class Response:  # private or not available at runtime\n"
"    code: int\n"
"    def get_header(self, name: str) -> str: ...\n"
"\n"
"def fetch_response() -> Response: ..."
msgstr ""
"@type_check_only\n"
"class Response:  # privada o no disponible en runtime\n"
"    code: int\n"
"    def get_header(self, name: str) -> str: ...\n"
"\n"
"def fetch_response() -> Response: ..."

Produces a

<rst-doc>:3: (ERROR/3) Unexpected indentation. while parsing: @type_check_only
class Response:  # privada o no disponible en runtime
    code: int
    def get_header(self, name: str) -> str: ...

def fetch_response() -> Response: ...

Marking those entries as fuzzy or adding three back-ticks "```" doesn't work, so I'm trying to brainstorm a bit how we could solve this.
A couple of ideas would be to:

1. Just try to fix at the docutils level (which is the source of the parsing error)
2. Introduce a new identifier like python-format in order to take the content for granted and avoid parsing it.

Ideas?

CC: @JulienPalard

使用 Doc/library/typing.rst 中引用的多行 PO 項目重現此故障，然後檢查 pospell 如何將程式碼項目傳遞給 docutils。確定所需的變更應位於 pospell 中，還是應在 docutils 層級進行；完成的標準應是這些項目能在不出現所回報的剖析錯誤的情況下獲得處理。