Improve SSTable codebase

[jaeyeom]

This commit introduces several improvements to the SSTable implementation:

• README.md: I enhanced this with details on SSTables, features, usage examples, build/test instructions, and contribution guidelines.
• Go Version: I updated the Go version in the GitHub Actions workflow to 1.21.
• Error Handling:
  • I replaced panic("unimplemented") calls with proper error returns.
  • I improved error messages to be more descriptive and contextual.
  • I wrapped underlying errors for better diagnostics.
• Code Clarity:
  • I refactored and clarified the noCursor logic for single-scan io.Reader.
  • I added extensive code comments to go/sstable/sstable.go for better readability and maintainability, covering package, types, functions, and internal logic.

These changes make your codebase more robust, user-friendly, and easier to understand and maintain.

[Copilot]

Pull Request Overview

This PR enhances the SSTable implementation by improving error handling, refactoring scan logic for different reader types, and expanding project documentation.

• Wrapped and contextualized errors in NewSSTable, replaced panics with proper returns, and introduced an errorCursor for scan errors.
• Refactored ScanFrom to handle io.ReaderAt, io.ReadSeeker, and plain io.Reader more clearly, with detailed comments around the noCursor logic.
• Expanded README.md with feature descriptions, usage examples, build/test instructions, contribution guidelines; updated Go version to 1.21 in GitHub Actions.

Reviewed Changes

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

File	Description
go/sstable/sstable.go	Improved error messages, replaced panics, added errorCursor, clarified scan logic and comments
README.md	Added detailed SSTable overview, usage examples, build/test/contribution instructions
.github/workflows/go.yml	Bumped Go version in CI from 1.13 to 1.21

Comments suppressed due to low confidence (1)

go/sstable/sstable.go:201

• The new errorCursor path for non-seekable readers and unsupported reader types isn't covered by existing tests; consider adding unit tests to verify error handling in these scenarios.

return &errorCursor{err: fmt.Errorf("ScanFrom: cursor already obtained for non-seekable reader type %T", r)}

[Copilot]

The inline comment notes a retained original behaviour without a clear TODO or issue reference; consider adding a formal TODO or implementing a proper error return for sequential scans with a non-nil key on an io.Reader.

Suggested change

	// It's not ideal to scan from beginning for a key with plain io.Reader without index.
	// However, the original logic attempts this.
	// A better approach might be to return an error if key is not nil and it's a plain io.Reader.
	// For now, maintaining existing behavior.
	// Returning an error for a non-nil key with a plain io.Reader without an index.
	// This avoids inefficient scanning from the beginning and makes the behavior explicit.
	return &errorCursor{err: fmt.Errorf("ScanFrom: cannot scan for a specific key with a non-seekable reader type %T without an index", r)}

[Copilot]

[nitpick] Scanning an entire SSTable sequentially to find a specific key without an index can be highly inefficient; it may be preferable to return an error or require a seekable reader when a non-nil key is provided.

Suggested change

	// It's not ideal to scan from beginning for a key with plain io.Reader without index.
	// However, the original logic attempts this.
	// A better approach might be to return an error if key is not nil and it's a plain io.Reader.
	// For now, maintaining existing behavior.
	// Return an error if a key is provided and the reader is non-seekable without an index.
	return &errorCursor{err: fmt.Errorf("ScanFrom: cannot scan for key %v with non-seekable reader and no index", key)}