fix: honor SSE-C chunk offsets in decryption for large chunked uploads (#8216)

* fix: honor SSE-C chunk offsets in decryption for large chunked uploads

Fixes issue #8215 where SSE-C decryption for large objects could corrupt
data by ignoring per-chunk PartOffset values.

Changes:
- Add TestSSECLargeObjectChunkReassembly unit test to verify correct
  decryption of 19MB object split into 8MB chunks using PartOffset
- Update decryptSSECChunkView and createMultipartSSECDecryptedReaderDirect
  to extract PartOffset from SSE-C metadata and pass to
  CreateSSECDecryptedReaderWithOffset for offset-aware decryption
- Fix createCTRStreamWithOffset to use calculateIVWithOffset for proper
  block-aligned counter advancement, matching SSE-KMS/S3 behavior
- Update comments to clarify SSE-C IV handling uses per-chunk offsets
  (unlike base IV approach used by KMS/S3)

All tests pass: go test ./weed/s3api ✓

* fix: close chunkReader on error paths in createMultipartSSECDecryptedReader

Address resource leak issue reported in PR #8216: ensure chunkReader is
properly closed before returning on all error paths, including:
- DeserializeSSECMetadata failures
- IV decoding errors
- Invalid PartOffset values
- SSE-C reader creation failures
- Missing per-chunk metadata

This prevents leaking network connections and file handles during
SSE-C multipart decryption error scenarios.

* docs: clarify SSE-C IV handling in decryptSSECChunkView comment

Replace misleading warning 'Do NOT call calculateIVWithOffset' with
accurate explanation that:
- CreateSSECDecryptedReaderWithOffset internally uses calculateIVWithOffset
  to advance the CTR counter to reach PartOffset
- calculateIVWithOffset is applied only to the per-part IV, NOT to derive
  a global base IV for all parts
- This differs fundamentally from SSE-KMS/SSE-S3 which use base IV +
  calculateIVWithOffset(ChunkOffset)

This clarifies the IV advancement mechanism while contrasting it with
the base IV approach used by other encryption schemes.

weed/s3api/s3_sse_c.go

@@ -263,18 +263,13 @@ func CreateSSECDecryptedReaderWithOffset(r io.Reader, customerKey *SSECustomerKe
 
 // createCTRStreamWithOffset creates a CTR stream positioned at a specific counter offset
 func createCTRStreamWithOffset(block cipher.Block, iv []byte, counterOffset uint64) cipher.Stream {
-	// Create a copy of the IV to avoid modifying the original
-	offsetIV := make([]byte, len(iv))
-	copy(offsetIV, iv)
-
-	// Calculate the counter offset in blocks (AES block size is 16 bytes)
-	blockOffset := counterOffset / 16
-
-	// Add the block offset to the counter portion of the IV
-	// In AES-CTR, the last 8 bytes of the IV are typically used as the counter
-	addCounterToIV(offsetIV, blockOffset)
-
-	return cipher.NewCTR(block, offsetIV)
+	adjustedIV, skip := calculateIVWithOffset(iv, int64(counterOffset))
+	stream := cipher.NewCTR(block, adjustedIV)
+	if skip > 0 {
+		dummy := make([]byte, skip)
+		stream.XORKeyStream(dummy, dummy)
+	}
+	return stream
 }
 
 // addCounterToIV adds a counter value to the IV (treating last 8 bytes as big-endian counter)