Handle attachment-backed Gmail body parts and declared charsets

[coderabbitai]

Summary

Improve Gmail body extraction so text and HTML message parts are handled correctly when their content is attachment-backed and when they declare a non-UTF-8 charset.

Today, the shared _extract_message_bodies helper in gmail/gmail_tools.py only consumes inline body.data and decodes it as UTF-8 with ignored errors. As a result, text/HTML body parts exposed through body.attachmentId can be omitted, and content in declared non-UTF-8 charsets can be corrupted in tools that rely on this helper.

Required changes

• Update the shared Gmail body extraction path to resolve body parts that provide body.attachmentId through the Gmail attachments API.
• Decode text/plain and text/html payloads using the charset declared by the MIME part when available, with safe fallback behavior for missing or invalid charset declarations.
• Preserve existing inline body.data extraction behavior.
• Apply the shared behavior consistently to:
  • get_gmail_message_content
  • get_gmail_messages_content_batch
  • get_gmail_thread_content
  • get_gmail_threads_content_batch
  • get_gmail_message_full for deliver_as="html" and deliver_as="txt"

Rationale

This is inherited shared-helper behavior rather than a narrow full-export concern. Fixing it in one focused change allows the contract and test coverage to be reviewed consistently across every Gmail body-reading tool.

get_gmail_message_full(deliver_as="eml") remains the byte-exact, lossless path because it exports Gmail's raw RFC 5322 bytes directly. The HTML/TXT modes are convenience representations and should nevertheless correctly retrieve body content and honor declared character encodings.

Acceptance criteria

• Attachment-backed text/plain and text/html body parts are retrieved and surfaced by all affected tools.
• Inline body.data behavior remains supported.
• Declared MIME charsets are honored for text/plain and text/html parts, with documented safe fallbacks.
• Tests cover attachment-backed body parts and at least one non-UTF-8 charset.
• Existing body-format and full-export tests continue to pass.

Backlinks

• Requested by @123andy
• PR feat(gmail): add get_gmail_message_full for untruncated body export #939: feat(gmail): add get_gmail_message_full for untruncated body export #939
• Review discussion: feat(gmail): add get_gmail_message_full for untruncated body export #939 (comment)