Adds ignore_unmapped option to nested and P/C queries

[colings86]

The change adds a new option to the nested, has_parent, has_children and parent_id queries: ignore_unmapped. If this option is set to false, the toQuery method on the QueryBuilder will throw an exception if the type/path specified in the query is unmapped. If the option is set to true, the toQuery method on the QueryBuilder will return a MatchNoDocsQuery. The default value is falseso the queries work how they do today (throwing an exception on unmapped paths/types)

[jpountz]

Can you use expectThrows instead? I think we should use it rather than this try:fail/catch:check_msg pattern for new code.

[jpountz]

A minor concern I have is that ignore_unmapped: false is a double negation, which could get confusing. Otherwise the change looks good and well tested.

[martijnvg]

maybe we should name this setting fail_unmapped? (and then default to true)

[colings86]

I'm ok with changing it to fail_unmapped although we seem to use ignore_* elsewhere in the api including ignore_malformed which is a bit of a double negation too. What do you think @jpountz and @clintongormley ?

[clintongormley]

ignore_unmapped defaults to false, so the only time the user will use this will be to set ignore_unmapped: true, which I think is OK.

[clintongormley]

Docs LGTM

[jpountz]

let's put it on the next line to be consistent?

[jpountz]

LGTM