fix(runtime): ask marimo for SameSite=None when the kernel is framed cross-site

[0x0f0f0f]

This pull request was authored by a coding agent.

Closes #328.

Under subdomain exposure the kernel is framed cross-site, so marimo's session cookie is third-party. Firefox drops it by default, the token exchange never persists, and marimo 303s to /auth/login — a page it serves with X-Frame-Options: DENY. Permanently blank frame.

Where the fix goes: the exposure is what knows the kernel is cross-site, so ExposurePreparation carries it the same way it already carries baseUrl for proxy. SubdomainExposure.prepare() asks for SameSite=None; the provisioner passes MARIMO_SESSION_COOKIE_SAMESITE. proxy is same-origin and asks for nothing.

Blocked on the matching marimo change that reads the riable. Older kernels ignore it and keep lax, so this is safe on a mixed fleet of sandbox images.

[vercel]

@0x0f0f0f-agents is attempting to deploy a commit to the marimo Team on Vercel.

A member of the Team first needs to authorize it.

[github-actions]

Caution

Pull Request opener is not an author or co-author of any commit in this PR.

• Opener: @0x0f0f0f
• Commit authors: @0x0f0f0f-agents

This check is blocked to guard against commits being submitted under a trusted identity the submitter does not control. If this PR is a legitimate cherry-pick, release-engineering submission, or mailing-list-style patch delivery, the repository maintainer can opt out of this check by setting require-opener-as-author: 'false' on the CLA-assistant step in the repository's workflow.

All contributors have signed the CLA ✍️ ✅
_{Posted by the CLA Assistant Lite bot.}

[0x0f0f0f-agents]

I have read the CLA Document and I hereby sign the CLA

[mscolnick]

Browsers can still block iframed cookies, even if SameSite=None. I will look into a solution and release

[mscolnick]

I am going to make this configurable (default off, like before). Generally, most sandbox providers can either provide their own ingress auth or provide some security through the entropy of the URL.

PR here: #330

Cookies as-is won't work well (safari and firefox still may block), but I will look to add support for Cookies Having Independent Partitioned State (CHIPS): https://developer.mozilla.org/en-US/docs/Web/Privacy/Guides/Third-party_cookies/Partitioned_cookies

[mscolnick]

@0x0f0f0f, ive released 0.4.2 that should be fixed. and ill make some more changes related to this in the next release