Merge lark sheets development branch

[zhengzhijiej-tech]

Summary

• Merge the lark sheets development branch into main.
• Includes the current sheets CLI updates from feat/lark-sheets-develop.

Testing

• Not run; merge PR only.

Summary by CodeRabbit

• New Features
  • Added Sheets shortcuts: +revision-get, +changeset-get, +formula-verify, and +history-list/+history-revert/+history-revert-status.
  • Enhanced +rows-resize / +cols-resize with --height/--width, --heights/--widths map form, plus font-family support and expanded chart gradients.
• Bug Fixes
  • Updated sheets --help to hide deprecated aliases from listings while keeping them executable.
  • Strengthened validation and error hints, including batch/resize and fan-out size/range caps; improved image token compatibility and +csv-put non-interactive stdin fallback.
• Documentation
  • Updated Sheets guides/references for the new shortcuts and revised resize/flag usage.

[zhengzhijiej-tech]

Follow-up attribution notes from the review:

• @xiongyuanwen-byted +batch-update sub-ops currently bypass the standalone cobra enum normalization/validation for flat flag-defs enum flags. Schema-backed enums are covered, but flat flags such as +rows-resize / +cols-resize --type can still enter the batch translator through mapFlagView without the chainEnumNormalization gate. Please add enum validation/normalization to the batch sub-op validation path or otherwise keep batch/standalone parity.
• @xiongyuanwen-byted the --size -> --height / --width surface change in +rows-resize / +cols-resize is intentional if the new CLI contract is allowed to break compatibility. If old commands must keep working, this needs a compatibility alias at the spec/CLI surface.
• @wenzhuozhen lark_sheet_history_revert.go has a stale file header: it says both shortcuts target --history-version-id, but +history-revert-status now correctly uses --transaction-id.
• @xiongyuanwen-byted validateDropdownRanges is now also used by +cells-batch-set-style; behavior looks fine, but the name is narrower than its current responsibility.

I opened #1839 only for the +changeset-get issue introduced by zhengzhijiej-tech.

[fangshuyu-768]

I found a few issues that look worth fixing before merging:

1. +batch-update sub-operations skip enum validation.

   In translateBatchOp, the batch path builds a mapFlagView and only calls validateRawTypes, which checks numeric and boolean fields but intentionally leaves string fields permissive. Standalone shortcuts still get enum validation from Cobra/common runner, so the same bad value is rejected standalone but can be accepted in a batch sub-op.

   For example, a +cells-clear sub-op with an invalid scope falls through normalizeClearType and becomes contents, and a +range-copy sub-op with an invalid paste-type falls through to all. That can turn a typo into a different write operation. The batch translator should validate enum-bearing string flags from flag-defs before calling the shortcut input builder.

2. +revision-get does not resolve wiki URLs during execution.

   RevisionGet.Execute calls resolveSpreadsheetToken, while the other sheets execute paths call resolveSpreadsheetTokenExec. For a /wiki/<node> URL, this sends the wiki node token as excel_id to get_workbook_structure instead of resolving it to the backing spreadsheet token first. This should use the Exec resolver, matching +workbook-info and the rest of the sheets commands.

3. +formula-verify --exit-on-error exits successfully for partial.

   formulaVerifyExitOnError treats status="partial" the same as success, but the new reference docs say partial means the scan was truncated and must not be treated as a clean verification. In CI or agent self-check flows, this can report success while part of the workbook was never scanned. With --exit-on-error, partial should return a non-zero typed error or otherwise force the caller to narrow/split the scan and retry until status="success".

4. Resize map operations are nondeterministic for overlapping ranges with the same start.

   resizeMapInput sorts generated operations only by the parsed start index. Since the input is a Go map, overlapping keys with the same start, such as {"A:C":100,"A:F":200}, can be emitted in either order. That changes the final width/height for the overlapping cells. Please add a deterministic tiebreaker, or reject overlapping ranges explicitly.

I also ran:

go test ./shortcuts/sheets/...
go test ./shortcuts/common ./shortcuts

[zhengzhijiej-tech]

@fangshuyu-768 Follow-up on items 1 and 3:

1. The batch enum-validation gap is real, but it was not introduced by this PR. At the review base (74d84586), translateBatchOp already called only validateRawTypes() before mapping.translate, so flat string enums were already bypassing standalone normalization. We still fixed the pre-existing gap in d8cffa94 by adding normalizeAndValidateEnums() to the batch path, with contract tests for invalid values, canonical casing, aliases, and hyphen/underscore keys.

2. The partial exit-code finding does not match the declared --exit-on-error contract. Both flag-defs.json and the command reference define this flag as returning non-zero only for status=errors_found, and TestFormulaVerifyExitOnError_StatusMatrix explicitly locks partial to a zero exit. partial is still emitted in the JSON with has_more=true, and the workflow docs require the caller to narrow/split the scan until status=success; it is not converted to success. Making partial non-zero would be a separate contract change rather than a correction to the current implementation, so we are keeping the existing behavior.