Minor improvements to publishing

[marta-lokhova]

Small follow up to #5081 to better handle some edge cases

• Move the exception about potentially corrupt checkpoints further down the recovery function to capture all possible recovery failures.
• Handle cases where publishing was enabled mid-checkpoint: skip publishing of the incomplete file, start publishing on the next checkpoint.

[Copilot]

Pull request overview

This PR improves checkpoint publishing robustness by handling edge cases where publishing is enabled mid-checkpoint. The changes refactor how incomplete checkpoints are detected and skipped, and relocate the corruption check to after the truncation process completes.

Key changes:

• Introduced mPublishWasDisabled flag to track when publishing was previously disabled and skip incomplete checkpoints
• Moved checkpoint file corruption detection to after truncation completes, avoiding false positives
• Changed corruption check from != to < to accommodate truncation scenarios
• Enhanced error messages when checkpoint files are partially missing

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 2 comments.

File	Description
src/history/CheckpointBuilder.h	Added publishWasDisabled() getter method and changed ensureOpen() return type from bool to void
src/history/CheckpointBuilder.cpp	Refactored file recovery logic in cleanup(), moved corruption check after truncation, updated appendTransactionSet() and appendLedgerHeader() to check publish status, and added validation for partial checkpoint file existence
src/history/HistoryManagerImpl.cpp	Added check to skip queueing incomplete checkpoints when publishing was previously disabled

[Copilot]

Pull request overview

Copilot reviewed 3 out of 3 changed files in this pull request and generated 2 comments.

[graydon]

I'm happy to see this getting more edge cases filed off, but I find it a little hard to follow I think mainly because of .. naming? Like it's a bit double-negative heavy: we have a flag tracking if the publication is disabled and then it's set to false once we start filling in a file. I think it might read better if we changed the variable to tracking whether we are currently open-and-appending, and then we return early if we're not at a checkpoint boundary and that's false. WDYT?