…us time index

This method would work like this:

|-- TIME_POINTS[0]
|
|             |-- TIME_POINTS[1]
|             |
|             ||-- TIMEPOINT_END_MARGIN
|             ||
|             ||   |-- DRAWWING_TIME EXTENSION
|             ||   |
|             ||   |       |-- Void time (annoucements, the show is going...)
|             ||   |       |
v  index: x   vv   v       v          index: y                      index: z
|=============|-|******|         |=============|-|******|         |=============|-|
              |++++++++++++++++++++++++++++++++|
                                               |________________________________|

In the range of |+|, `get_prev_time_index` will return `x`.
In the range of |_|, `get_prev_time_index` will return `y`.

…commit

The diagram on the commit message was wrong,
so DON'T PREFER TO IT.
Instead, see notes.
you can show notes by using:
```
git notes show 3fb5c1a
```

…dex'

Test whether it fail when the time is out of acceptable range

…e_index'

…e edge of range

If I do `ends[i] <= time <= ends[i+1]`,
This expression will be True twice when `time` is the same as
`ends[i+1]` because that would be the same as `ends[i] for the
next expression.

Of cource, this conflict won't work because of looping,
But it is not good idea to leave things confusing, right?
So I fixed it

…me_index'

'data' will be thrown to the jinja2 template

…on 'kind'

it's because the way generating PDF could be hard to do in such a
limited time.
I found that wkhtmltopdf doesn't treat CSS on alpine.
It'll take more times to solve

…ture/183-generate-media-to-announce