Execute code and shell commands inside a Daytona sandbox with correct timeouts, exit codes, env vars and working directory

Documented steps

1. For one-off, stateless execution call Python `sandbox.process.code_run(code, params=CodeRunParams(argv=[...], env={...}), timeout=...)` or TypeScript `sandbox.process.codeRun(code, params, timeout)`.
2. For stateful, notebook-style execution that keeps variables between calls, use Python `sandbox.code_interpreter.run_code(code, context=..., envs={...}, timeout=600, on_stdout=..., on_stderr=..., on_error=...)` or TypeScript `sandbox.codeInterpreter.runCode(...)`.
3. Isolate interpreter state with `create_context(cwd)`, enumerate with `list_contexts()`, and release with `delete_context(ctx)`.
4. For arbitrary shell commands call Python `sandbox.process.exec(command, cwd='workspace/src', env={...}, timeout=30)` or TypeScript `sandbox.process.executeCommand(command, cwd, env, timeout)`.
5. Always pass an explicit `timeout` to `exec`/`executeCommand`: the default timeout is 10 seconds when not specified.
6. After every call read `result` for combined output and check `exit_code` numerically to decide success — do not rely on exceptions alone.
7. Supply command-line arguments and environment variables through `CodeRunParams(argv=[...], env={...})` rather than string-interpolating them into the code, which avoids quoting bugs.
8. When the executed code produces matplotlib-style plots, read the returned `artifacts.charts` array to retrieve chart output.
9. Wrap calls in try/except for `DaytonaError` (Python) or the equivalent SDK exception, and treat a timeout as a distinct outcome from a non-zero exit code.
10. Set `timeout=0` on `run_code` only when you genuinely want no limit, and pair it with an external watchdog. Official docs: https://www.daytona.io/docs/en/process-code-execution