Fix GzipDecompressor to handle concatenated gzip members

[veeceey]

Summary

Fixes #3560

The GzipDecompressor class now properly handles concatenated gzip members by checking for unused_data after each decompression and creating a new decompressor to continue processing subsequent members. This prevents silent data loss when decompressing multi-member gzip streams, which is a common pattern in HTTP streaming and other applications.

Changes

• Modified GzipDecompressor.decompress() to loop through concatenated gzip members using unused_data
• Added _flushed flag to prevent decompress() calls after flush() has been called
• Added comprehensive test cases covering:
  • Single gzip member (regression test)
  • Two concatenated gzip members (issue reproduction)
  • Three concatenated gzip members (edge case)
  • RuntimeError after flush() (API contract)

Test Results

All existing tests pass, and new tests verify the fix:

$ python3 -m pytest tornado/test/util_test.py::GzipDecompressorTest -v
============================== test session starts ==============================
tornado/test/util_test.py::GzipDecompressorTest::test_concatenated_gzip_members PASSED
tornado/test/util_test.py::GzipDecompressorTest::test_decompress_after_flush_raises PASSED
tornado/test/util_test.py::GzipDecompressorTest::test_multiple_concatenated_members PASSED
tornado/test/util_test.py::GzipDecompressorTest::test_single_gzip_member PASSED
============================== 4 passed in 0.02s

The reproduction script from the issue now works correctly:

import gzip
from tornado.util import GzipDecompressor

data1 = b"This is some example data that will be compressed using gzip."
data2 = b"Here is some more example data to demonstrate gzip compression."

member1 = gzip.compress(data1)
member2 = gzip.compress(data2)
concatenated = member1 + member2

decompressor = GzipDecompressor()
decompressed_data = decompressor.decompress(concatenated)
expected_data = data1 + data2

assert decompressed_data == expected_data  # ✓ Passes with fix