Run CLI E2E tests from portable command and golden files

[ColinLeeo]

CLI end-to-end tests currently call run_cli() inside Google Test and depend on the C++ source/build tree. They cannot be reused directly to validate a separately installed CLI.

Move executable-level scenarios into a standalone cpp/test/tools/e2e/ bundle containing cases.json, fixtures, goldens, and a Python standard-library runner. The runner invokes the specified executable, checks exit status independently from stdout/stderr and exported files, preserves multi-command write/read scenarios, and retains actual outputs, exit codes, diffs, and identity hashes. Existing parser and injected metadata/stream failure tests remain C++ unit tests.

The bundle has 155 cases and 1,076 command invocations, preserves the original 141 goldens, and adds read-back checks for successful writes plus a redirected-stdin round trip. CTest runs the same bundle against the build-tree CLI; CI uploads failure results. Python 3.9+ is required when tools and tests are enabled. Product implementation code is unchanged.

Validation on macOS ARM64:

• Original 170 tests passed with their existing assertions before migration.
• CMake Release build of TsFile_Test passed; 42 retained CLI/parser/format/statistics unit tests passed.
• CliE2E and CliE2ERunnerTest passed: all 155 cases / 1,076 commands and 14 runner regression tests.
• Exported the standalone bundle from the Git index and reran all cases successfully outside the checkout.
• POSIX-step skip behavior was also exercised on macOS; this is not a Windows execution result. Linux and Windows validation remains for PR CI.
• Apache RAT, Black 26.3.1, clang-format 17, and non-fixture whitespace checks passed.

Usage:

python3 cpp/test/tools/e2e/run_cli_e2e.py \
  --cli /path/to/installed/tsfile-cli \
  --output-dir /tmp/tsfile-cli-results