Several improvements to token filter types

[JoshMock]

An issue on the JS repo was opened about a missing cjk_bigram token filter type. I did an audit of the token filter docs and found a few missing types, so I:

• Added the missing types
• Added some missing values to enums
• Moved a couple types into their relevant plugin files
• Created a few base types for token filters that share most of the same parameters
• Added docstrings to as many parameters as I could find
• Removed a few parameters not mentioned in the docs

Fixes #4251
Fixes elastic/elasticsearch-js#2726

[l-trotta]

@JoshMock I'm having some doubts about this one, I think while it's good to have the full list of stopwords, some of the specific tokenizers only accept part of it? For example, this is the code for the JapaneseStopTokenFilter, where stopwords seems to be a list of strings defined in Lucene

[JoshMock]

@l-trotta Good catch. I updated the stopwords types in 5fc7dbe.

[flobernd]

Thank you @JoshMock 🙂 Just a minor comment.

[flobernd]

This will cause inconveniences for users in statically typed languages.

Could we please continue using the StopWords type here?

StopWords is currently defined as:

export type StopWords = string | string[]

Changing that to:

export type StopWords = StopWordLanguage | string[]

while keeping this file "as is", should do the trick.

What do you think?

[l-trotta]

I didn't comment on this because the java client simplifies enum | string to string, so I assumed other static clients would have something similar

[flobernd]

For .NET I don’t do this since enums are way nicer to use.

In this case it’s as well string[]. Do you simplify that as well?

[l-trotta]

yeah it also works as enum | string[] -> string[]

[JoshMock]

@flobernd I think I got what you mean. Made a change in db8c130. Let me know if that looks better to you.

[flobernd]

Perfect now! Thank you @JoshMock