Refactor geo_point validate* and normalize* for 2.x

[nknize]

This PR refactors the old validate and normalize options for geo_point field type to ignore_malformed and coerce, respectively. Geo queries are also updated to ignore_malformed GeoPoints. This PR is for 2.0 only since mapping implementation has drastically changed from 1.x. For 1.7 changes see PR #12300

closes #10170

[colings86]

I like this change but it does mean that the meaning of coerce here would be subtly different from the meaning in other types. For example if I have an integer field and I set coerce: false and then try to index a document where the value of that field is "200" then it will produce and error (if ignore_malformed is also false). with this change (if I am reading correctly), a geo_point field with coerce: false (and ignore_malformed: false) will not error if I insert a geo_point [480.5, 160.9] but will index the geo_point with the un-normalised lat and long values.

[nknize]

@colings86 The same applies. (e.g., see GeoPointFieldMapperTests.testValidateLatLonValues) If you set ignore_malformed: false and coerce: false and insert a malformed geo_point (e.g., coordinate [1.3, -91]) the parser will throw an Exception.

[colings86]

oh ok, great. Thanks for clarifying :)

[rjernst]

I don't like this bwc member. It is non-descript, and requires jumping to where it is set to figure out under what version this setting is no longer allowed. I think we should put the version check with the setting parsing.

[rjernst]

You could also just name it, as a local, so that the name implies what version is being checked for, but I would personally just put the version check directly here.

[rjernst]

Actually now that I see the rest of how this works, I think we should just be checking version here, and set the equivalent setting using ignoreMalformed/coerce. You can also add the version check into the conditional next to the name check, and let it fall through on newer indexes (this will cause an unknown setting exception). All of the old names can then be removed from the fieldtype/builder/mapper.

[rjernst]

I think you need to make this needs to be change, it can't be beta1 anymore?

[rjernst]

@nknize I left a couple more comments. Mappings look good, I think we should add some more tests.

[rjernst]

I find the name a little confusing, since we already have this exact same name in FieldMapper, and it means before beta1. A better name would be good...I don't have a suggestion though...

[rjernst]

This looks great. I left couple more comments about the tests to make sure we have real backcompat checks there.

[nknize]

Thanks for the review @rjernst! Updated bwc testing and other minor edits from the comments.

[rjernst]

LGTM, 1 more minor suggestion for the tests.