Xidorn Quan

To support this, we may need to fix #29 first, at least so that we can handle non-ASCII content for description clipping. This would probably make #30 unnecessary. Also, #32...

> We may also want to save generated fst into files so that we don't need to re-generate them every time the bot restarts. This is now #35.

I have a feeling that it might be fine to just support Serde by treating all the types as the next pow of 2 primitive type (i.e. `u7` -> `u8`,...

I don't think changing the internal representation would affect serialization as long as there is still way to convert between primitives and the types.

This is a duplicate of #73

@glennw suggested in IRC that it may be reasonable to scale the delay time based on the priority, so that higher priority PR has longer waiting time. And @mbrubeck suggested...

问叹号在GB/T 15834-2011中实际上是有记录的，作为问号和叹号的结合形式（见4.3.3.4）。因此我怀疑该符号不应视为一个独立符号。不过问号和叹号连用的这种形式，至少应在文档中有所体现。 不过如果文档里已经有双叹号和双问号的记载，或许作为独立的符号也没有问题，但我个人理解中还是倾向于把它们全部当作在标点挤压支持不完善时的workaround。

实际上你说的前四种都存在，其中双叹号的U+203C原本就是为了workaround直排下的双叹号。 另外根据现有的标准做法， 连续的标点在横排时确实不会被切断。至于直排时视为一个字元的问题，日文对两位数字也有类似的需求，Writing Modes标准引入了text-combine-upright属性来支持。这个问题或许可以使用类似的方式解决。

而且正常来说，一般人并不会输入这几个特殊符号，而是采用连续的单个符号来替代。

Yeah, we should still gate the API behind user activation.