feat(models_ollama): Include model capabilities

[gadenbuie]

I'm on a plane and didn't have WiFi, which reminded me that one bit of friction with local models via Ollama is that I can never remember which models have tool calling capabilities.

It turns out that you can query a model's details via the /api/show endpoint. I thought this would be useful information to have, so I used this to add a capabilities column to models_ollama(). Having to query one-at-a-time slows down the call to models_ollama() at about 10ms per model (on my Macbook Air M3), but we can cache model details to avoid paying for the model details more than once.

models_ollama() |> dplyr::as_tibble()
#> # A tibble: 10 × 4
#>    id                              created_at                size capabilities             
#>    <chr>                           <dttm>                   <dbl> <chr>                    
#>  1 qwen3:14b                       2025-07-05 00:00:00 9276198565 completion,tools,thinking
#>  2 phi4-mini                       2025-07-05 00:00:00 2491876774 completion,tools         
#>  3 granite3.3:8b                   2025-07-03 00:00:00 4942891653 completion,tools         
#>  4 gemma3:4b                       2025-07-03 00:00:00 3338801804 completion,vision        
#>  5 qwen3:4b                        2025-05-12 00:00:00 2620788019 completion,tools         
#>  6 qwen2.5-coder:14b-instruct-q2_K 2025-03-02 00:00:00 5770511454 completion,tools,insert  
#>  7 deepseek-r1:14b                 2025-01-30 00:00:00 8988112040 completion               
#>  8 nomic-embed-text                2024-10-14 00:00:00  274302450 embedding                
#>  9 llama3.1:8b                     2024-10-08 00:00:00 4661230766 completion,tools         
#> 10 llama3.2                        2024-10-01 00:00:00 2019393189 completion,tools