Inspection of arguments in function cause inspection of surrounding function

At the moment inspection of arguments in function calls is not possible as the function is inspected instead. This is because `do_inspect` relies on `IPython.utils.tokenutil.generate_tokens`:

https://github.com/ipython/ipykernel/blob/c24b252dc4fc81e6cf6354f8d64ea06ed1ce3496/ipykernel/ipkernel.py#L544-L546

The behaviour of `generate_tokens` has changed in https://github.com/ipython/ipython/commit/240eef90ab769a25100eb9f2c4247d775e6c160c . Before it would return the tokens of names, including parameters and names, even if they were inside a function call. Not all edge cases were handled correctly, but for most day-to-day inspection tasks were served fine. Now `generate_tokens` returns function names if the cursor is positioned inside the parentheses:

```
from IPython.utils.tokenutil import generate_tokens, token_at_cursor
from io import StringIO
code = 'func(param1, param2=arg)'
tokens = list(generate_tokens(StringIO(code).readline))

pos = dict(func=3, param1=8, param2=15, arg=22)
# Current behaviour
dict((k, token_at_cursor(code, v) for k, v in pos.items())
# Output: {'func': 'func', 'param1': 'func', 'param2': 'func', 'arg': 'func'}
```

Start in ipykernel/ipkernel.py around lines 544-546 and inspect how do_inspect uses generate_tokens, token_at_cursor, and the surrounding cursor handling. Reproduce the provided func(param1, param2=arg) example, then verify that cursors on arguments identify those arguments rather than the surrounding function.

由索引模型根据 Issue 内容生成。