Resolves three analyzer warnings that were causing CI build failures:

1. CA1868: Use HashSet.Add return value instead of Contains + Add pattern
   - Changed lines 230-231 in ListItemsAsync to use the boolean return
     value of HashSet.Add() directly, eliminating the redundant Contains check

2. CA1835: Use Memory-based Stream overloads (2 instances)
   - Changed lines 332-333 in ReadFileAsync to use ReadAsync(Memory<byte>)
     and WriteAsync(ReadOnlyMemory<byte>) instead of array-based overloads
   - These overloads are more efficient and are the recommended pattern
     for .NET 8.0 (our target framework)

Technical details:
- HashSet.Add returns false if item already exists, true if added
- Memory<T> overloads reduce allocations and improve performance
- Both changes maintain identical functionality while satisfying analyzers

Files changed:
- src/SharpSync/Storage/S3Storage.cs: Fixed 3 analyzer warnings

Resolves build failure in S3StorageTests where ToArray() was called on
a Stream object. The ToArray() method only exists on MemoryStream,
not on the base Stream class.

Change in ProgressChanged_LargeFileDownload_RaisesEvents test:
- Before: readStream.ToArray() (invalid - Stream has no ToArray method)
- After: Copy stream to MemoryStream using CopyToAsync()

This properly consumes the stream and triggers progress events as intended,
while fixing the compilation error CS1061.

Technical details:
- Stream is an abstract base class without ToArray()
- MemoryStream inherits from Stream and provides ToArray()
- CopyToAsync() fully reads the stream, triggering progress reporting
- Using statement ensures proper disposal of both streams

Files changed:
- tests/SharpSync.Tests/Storage/S3StorageTests.cs: Fixed line 550

Resolves failing integration tests where ListItemsAsync was not returning
file objects, only directory markers.

Root cause:
- S3 prefix was missing trailing slash when listing directories
- This caused S3 to treat the prefix as a string match rather than a
  directory boundary, returning incorrect results
- Objects were being incorrectly filtered out

Changes:
1. Add trailing slash to listPrefix for directory listings
   - Root listing: use "prefix/" if prefix exists, "" otherwise
   - Subdirectory listing: ensure "path/" format with trailing slash

2. Simplified object filtering
   - Only skip keys ending with '/' (directory markers)
   - Remove check for exact prefix match (already handled by trailing slash)

3. Improved comments explaining S3 prefix/delimiter behavior

Technical details:
- S3 ListObjectsV2 with Delimiter="/" returns:
  - S3Objects: actual file objects in this "directory"
  - CommonPrefixes: subdirectories (keys with trailing /)
- Prefix MUST end with '/' to list directory contents correctly
- Without trailing slash, S3 returns all keys starting with prefix string

This fixes tests:
- ListItemsAsync_WithFiles_ReturnsFiles (expected file1.txt, file2.txt, subdir)
- ListItemsAsync_Subdirectory_ReturnsOnlySubdirectoryContents (expected 2 files)

Files changed:
- src/SharpSync/Storage/S3Storage.cs: Fixed ListItemsAsync prefix handling