* feat(logging): integrate logger_system for centralized logging

Replace std::cout/cerr with structured logging throughout network_system.
This provides better debugging capabilities, log level filtering, and
prepares for future monitoring integration.

Changes:
- Add logger_integration interface with basic_logger and logger_system_adapter
- Replace initial std::cout/cerr calls with NETWORK_LOG macros
- Add BUILD_WITH_LOGGER_SYSTEM CMake option for conditional compilation
- Implement timestamp and source location tracking

* docs: update documentation for logger system integration

- Add logger_integration.h to architecture diagram in README
- Include BUILD_WITH_LOGGER_SYSTEM in build configuration examples
- Document logger system as optional dependency
- Add comprehensive logger integration section to CHANGELOG
- Create new INTEGRATION.md guide with detailed logger usage
- Document all integration options and configuration methods

* fix: correct header include paths in legacy files

Update include paths from "network/" to "network_system/" in legacy
compatibility files to maintain consistency with the new project structure.
This ensures proper header resolution when building the project.

Changed files in core/, session/, and internal/ directories to use
the correct project-wide header paths while maintaining network_module
namespace for backward compatibility.

* fix: resolve atomic log_level compilation issue on Linux

- Add explicit underlying type (int) to log_level enum class
- Use std::atomic<int> instead of std::atomic<log_level> for portability
- Add explicit casts when comparing/assigning log levels
- Include <atomic> header for std::atomic support

This fixes compilation errors on GCC/Linux where std::atomic
requires complete type information for enum classes.

* feat(integration): update thread_system and logger_system integration

- Update thread_system to use kcenon::thread namespace and new headers
- Fix messaging_bridge to use kcenon::thread::thread_pool instead of thread_system::thread_pool
- Add thread_system and logger_system include paths to all test targets
- Enable proper linking of updated thread_base and LoggerSystem libraries
- Remove obsolete Makefile in favor of CMake build system

All tests now build successfully with updated dependencies.

* chore(build): add test executables to .gitignore

- Ignore Unix/macOS test executables without extensions
- Prevent test binaries from being accidentally committed

* chore(build): remove test executables from git tracking

- Remove binary executables stress_test and test_e2e from version control
- These should be generated during build, not tracked in repository

Round-2 reviewer feedback (M1, Issue #1115): the Round-6 direct-call
refactor changed the semantics of ServerUnknownFrameTypeIsHandled-
WithoutCrashing — the original wire-level test sent a raw frame header
with type=0xFF, exercising EITHER the process_frame default switch arm
OR the run_io parse-error path (RFC 7540 §5.5 unknown-type discard).

Direct invocation through Http2ClientTestAccess::process_frame cannot
construct a frame instance with an undefined type because frame::parse
rejects unknown types upstream of process_frame. The TEST_F now passes
a null unique_ptr<frame>, which exercises the null-pointer defensive
guard at src/protocols/http2/http2_client.cpp:620 — a complementary
arm, not the unknown-type dispatch.

The unknown-type wire path remains covered by
MalformedServerSettingsTypeByteTriggersConnectTimeout (which lives on
the multiplier scaffold and uses frame_injector to corrupt the type
byte on a real socket).

Renamed in place (no new TEST_F added, AC #6 preserved):
- ServerUnknownFrameTypeIsHandledWithoutCrashing
+ ProcessFrameNullPointerGuardReturnsInvalidArgument

The implementation comment block is rewritten to explain the semantic
shift so future maintainers see why the rename happened.

Round-2 reviewer feedback (M1, Issue #1115): the Round-6 direct-call
refactor changed the semantics of ServerUnknownFrameTypeIsHandled-
WithoutCrashing — the original wire-level test sent a raw frame header
with type=0xFF, exercising EITHER the process_frame default switch arm
OR the run_io parse-error path (RFC 7540 §5.5 unknown-type discard).

Direct invocation through Http2ClientTestAccess::process_frame cannot
construct a frame instance with an undefined type because frame::parse
rejects unknown types upstream of process_frame. The TEST_F now passes
a null unique_ptr<frame>, which exercises the null-pointer defensive
guard at src/protocols/http2/http2_client.cpp:620 — a complementary
arm, not the unknown-type dispatch.

The unknown-type wire path remains covered by
MalformedServerSettingsTypeByteTriggersConnectTimeout (which lives on
the multiplier scaffold and uses frame_injector to corrupt the type
byte on a real socket).

Renamed in place (no new TEST_F added, AC #6 preserved):
- ServerUnknownFrameTypeIsHandledWithoutCrashing
+ ProcessFrameNullPointerGuardReturnsInvalidArgument

The implementation comment block is rewritten to explain the semantic
shift so future maintainers see why the rename happened.