Fix tar archive enumeration after fully reading entry streams

[adamhathcock]

master PR of #1324

This pull request refactors the handling of tar entry streams to ensure the underlying archive stream is always correctly positioned at the start of the next entry, even if entry streams are not explicitly disposed. It introduces new logic in TarReadOnlySubStream to automatically advance to the next header when reading is complete, and adds comprehensive tests to verify correct behavior when reading entries one at a time without disposing the entry streams.

Core improvements to tar entry stream handling:

• Refactored TarReadOnlySubStream to automatically advance the archive stream to the next header (including padding) when the end of an entry is reached, regardless of whether Dispose is called, by introducing AdvanceToNextHeader and AdvanceToNextHeaderAsync methods. This ensures robust and predictable behavior when consumers do not dispose entry streams. [1] [2] [3] [4] [5] [6] [7] [8] [9]

• Removed the useSyncOverAsyncDispose parameter from TarReadOnlySubStream and all related code, simplifying the stream's construction and disposal logic. [1] [2] [3] [4] [5]

Expanded test coverage:

• Added new tests in TarArchiveTests, TarArchiveAsyncTests, TarReaderTests, and TarReaderAsyncTests to verify that tar entries can be read one at a time without disposing the entry stream, and that all entries are still accessible and correct. [1] [2] [3] [4] [5] [6] [7]

These changes make tar archive reading more robust and user-friendly, preventing subtle bugs when entry streams are not disposed and ensuring correct archive navigation in all scenarios.

[Copilot]

Pull request overview

This PR fixes tar archive iteration when an entry stream is fully consumed but not immediately disposed, ensuring the underlying tar stream advances to the next header boundary.

Changes:

• Adds idempotent sync/async header-advance logic to TarReadOnlySubStream.
• Simplifies TarReadOnlySubStream construction by removing the sync-over-async constructor flag.
• Adds sync and async regression tests for archive and reader APIs.

Reviewed changes

Copilot reviewed 8 out of 8 changed files in this pull request and generated no comments.

Show a summary per file

File	Description
src/SharpCompress/Common/Tar/TarReadOnlySubStream.cs	Advances past remaining entry data and padding when reads reach EOF.
src/SharpCompress/Common/Tar/TarHeaderFactory.cs	Updates tar substream construction for sync streaming reads.
src/SharpCompress/Common/Tar/TarHeaderFactory.Async.cs	Updates tar substream construction for async streaming reads.
src/SharpCompress/Common/Tar/TarFilePart.cs	Updates seekable tar entry stream construction.
tests/SharpCompress.Test/Tar/TarArchiveTests.cs	Adds sync archive regression coverage.
tests/SharpCompress.Test/Tar/TarArchiveAsyncTests.cs	Adds async archive regression coverage.
tests/SharpCompress.Test/Tar/TarReaderTests.cs	Adds sync reader regression coverage.
tests/SharpCompress.Test/Tar/TarReaderAsyncTests.cs	Adds async reader regression coverage.