CLI: Reduce installation noise and improve error handling

[yannbf]

Closes #22343

What I did

Before, errors thrown by addDeps from our package manager proxies would always look something like this, without proper explanation:

Command failed with exit code 1: npm install -D storybook@^7.1.0-alpha.11 @storybook/react@^7.1.0-alpha.11 @storybook/react@^7.1.0-alpha.11 @storybook/react-vite@^7.1.0-alpha.11 @storybook/addon-links@^7.1.0-alpha.11 @storybook/addon-essentials@^7.1.0-alpha.11 @storybook/blocks@^7.1.0-alpha.11 @storybook/addon-interactions@^7.1.0-alpha.11 @storybook/testing-library@^0.0.14-next.2 prop-types@^15.8.1

In this PR, I do a few things:

• Add a log stream implementation to generate log files
• Use that implementation as part of all package manager proxies when installing dependencies. The log stream will reduce the installation noise by 100%, but I also added a "Installing dependencies" log for users to know what's going on.
• Upon failure, each package manager will try and proxy the errors and come up with better error messages
• For each package manager (but yarn1) there's a list of errors which will be matched, else it will show e.g. "Unknown PNPM error"
• Add ora package to display spinner in the CLI (now that users don't see progress they might end up quitting the CLI if we don't display a spinner)

Before:

After (example for each package manager):

And an example of how the log file looks like:

Spinners in action:

How to test

1. Build the Storybook CLI
2. Choose a project that does not have Storybook (use sandboxes repo) e.g. react-webpack
3. run STORYBOOK_DISABLE_TELEMETRY=true storybook init command (using the built binary) in those
4. Watch the experience

when testing the init command, you can also pass a package manager via the --package-manager flag

Checklist

• Make sure your changes are tested (stories and/or unit, integration, or end-to-end tests)
• Make sure to add/update documentation regarding your changes
• If you are deprecating/removing a feature, make sure to update
  MIGRATION.MD

Maintainers

• If this PR should be tested against many or all sandboxes,
  make sure to add the ci:merged or ci:daily GH label to it.
• Make sure this PR contains one of the labels below.

["cleanup", "BREAKING CHANGE", "feature request", "bug", "documentation", "maintenance", "dependencies", "other"]

[shilman]

This is awesome @yannbf -- was going to ask for this as a next step to the error handling project but you beat me to it. 😅

[shilman]

RE: UX can we show a CLI spinner while the installation is occurring now that we're swallowing all the installer output? Also we probably need to handle ctl+C behavior.

[yannbf]

TODO: Move process.on('SIGINT'...) to withTelemetry and send a canceled event, something like:

export async function withTelemetry<T>(
  eventType: EventType,
  options: TelemetryOptions,
  run: () => Promise<T>
): Promise<T> {
  process.on('SIGINT', async () => {
    if (!options.cliOptions.disableTelemetry) {
      await telemetry('canceled', { eventType }, { stripMetadata: true });
    }

    // figure out how to exit properly from here
    // allowing SIGINT default handling mechanism to run
  });

  if (!options.cliOptions.disableTelemetry)
    telemetry('boot', { eventType }, { stripMetadata: true });

  try {
    return await run();
  } catch (error) {
    const { printError = logger.error } = options;
    printError(error);

    await sendTelemetryError(error, eventType, options);
    throw error;
  }
}

And make sure this yields the correct result when users do Ctrl + C in the middle of init, rather than generate an error that is unrelated!