Parse a scene's summary as a real list of bullet points

A frontmatter `summary:` list written as a multi-line flow sequence
(one bullet per line, closing bracket on its own line — the natural
way to hand-write several sentences) used to vanish: no line in it
has a colon or a leading `-`, so the parser read none of the lines
after the opening `summary: [` and the field ended up holding a
literal "[". Even written on one line, `flow_list_items` split on
every comma, which cuts a prose bullet in two if it has one of its
own.

`Header` now has a `summary: Vec<String>`, read by a parser that
joins a flow sequence's continuation lines before splitting it on
commas outside of quotes, so each bullet survives whole. Also added
`summary: []` to the bundled scene template and the in-app default
new-file template, so new scenes offer the field.

No new UI surfaces this yet (corkboard/synopsis still show only the
slug) — out of scope for the parsing fix itself.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

templates/snowflake/{{ cookiecutter.project_name }}/05-First Draft/_template.md

@@ -11,6 +11,7 @@ status: draft
 pov: ""
 characters: []
 setting: ""
+summary: []
 date: ""
 time: ""
 conflict: ""