fix(ch06): guard against IndexError on empty LLM choices list

[qizwiz]

Problem

Nine example scripts in chapter 06 call choices[0].message.content without checking whether the API returned any choices. This raises a bare IndexError when:

• the LLM applies a content-policy filter and returns choices=[]
• a token limit is hit before the first choice is generated
• the API returns a non-standard error response body

These are tutorial scripts that learners copy and run — an IndexError here is confusing and makes the example look broken.

Changes

File	Fix
aoai-app.py	if completion.choices: before printing
aoai-history-bot.py	same
aoai-study-buddy.py	same
oai-app.py	same
oai-history-bot.py	same
oai-study-buddy.py	same
githubmodels-app.py	if response.choices: before printing
aoai-app-recipe.py	guard both choices accesses; skip dependent shopping-list call when first response is empty
oai-app-recipe.py	same

All changes are minimal defensive checks — no logic changes for successful API responses.

---

Found by pact static analysis — llm_response_unguarded mode.

[github-actions]

👋 Thanks for contributing @qizwiz! We will review the pull request and get back to you soon.

[qizwiz]

Update: extended guards to cover Gemini content-filter crash path

Testing against live providers found that Gemini 2.5 Flash returns HTTP 200 with choices[0].message = None (not an empty choices list) when content is filtered — finish_reason: PROHIBITED_CONTENT. The previous guards if completion.choices: only prevent IndexError on empty lists; AttributeError: 'NoneType' object has no attribute 'content' still occurs when message itself is None.

Pushed follow-up commit (05bd9ac) extending all 9 app files to the comprehensive guard:

if completion.choices and completion.choices[0].message is not None:
    print(completion.choices[0].message.content)

This covers both crash vectors in a single check.

[qizwiz]

Quick status summary for reviewers:

This PR guards against two distinct crash paths in the chapter 06 tutorial scripts:

1. Empty choices (IndexError) — when content policy filters the response, choices=[]. Fixed with if completion.choices: before indexing.
2. None message (AttributeError) — Gemini 2.5 Flash returns HTTP 200 with choices[0].message = None when finish_reason: PROHIBITED_CONTENT. Fixed with an additional and completion.choices[0].message is not None check.

Both are real failure modes learners will hit when running these examples against live APIs. The guards are minimal and don't change the tutorial logic — they just print a friendly message instead of crashing.