Fix enterprise account failures and add Org ID field (#940)

[clintandrewhall]

Summary

Fixes #940 — enterprise/organization Claude accounts always fail with "Invalid response from Claude API".

• Parser crash on null quota fields: Enterprise credit-based accounts return null for five_hour (and seven_day) instead of utilization data. The parser was throwing FetchError.invalidResponse when five_hour was absent/null. Now defaults sessionPercentUsed to 0.0 so enterprise accounts load correctly.
• Missing Org ID UI field: The preferences UI had no way to enter an organizationID for accounts linked to multiple Anthropic organizations, forcing users to manually edit JSON config. Added an optional "Org ID" text field to the Claude token account entry form, wired through to addTokenAccount(organizationID:).

Changes

• ClaudeWebAPIFetcher.swift — replace guard-throw with ?? 0.0 default for missing session utilization
• ProviderSettingsDescriptors.swift — add showsOrganizationField: Bool; update addAccount closure to accept optional organizationID
• PreferencesProviderSettingsRows.swift — show "Org ID (optional)" text field when showsOrganizationField is true; pass value to addAccount
• PreferencesProvidersPane.swift — enable org field for Claude provider; forward organizationID to addTokenAccount
• PreferencesProvidersPane+Testing.swift — update mock descriptor to match new signature

Test plan

• Add a Claude web session key account — verify Label, Token, and Org ID fields appear; submitting with a blank Org ID creates an account with no org override
• Add a Claude account with an Org ID — verify the value is saved and used when fetching usage
• With an enterprise Claude session key, verify usage fetches successfully (no "Invalid response" error) and sessionPercentUsed shows 0%
• Non-Claude token account providers (e.g. Copilot) should not show the Org ID field

https://claude.ai/code/session_01YQotcG4rFPbYNkyZLt94hr

---

Generated by Claude Code

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: cdb0a8f442

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 04a270f3da

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Preserve parse failure when session window is absent

parseUsageResponse now treats any missing/unparseable five_hour window as 0.0 usage, not just the enterprise null case. That means malformed or schema-shifted 200 responses can be accepted as success, and StepExecutor.executeAuto will not fall back to CLI because no error is thrown on the web step. In those cases users can get a false 0% session reading instead of the previous fallback behavior. Consider limiting the default-to-zero path to explicit five_hour: null and still throwing for absent/invalid window structures.

Useful? React with 👍 / 👎.

[clintandrewhall]

Thanks for the quick turnaround, @steipete -- I don't typically write in Swift, but I thought it was a simple enough fix that I could leave it to Claude. Hope it helped!