Fixes XXE vulnerability in Eureka Server (external entity processing
on by default in older versions of the library).

* Check if object is already a ResponseEntity, if so return instead of wrapping in another ResponseEnitiy. Which caused headers and status code from original ResponseEnitiy to be lost and for body to be output with whole ResponseEntity instead of just the body

* added test

Fixed an issue where the pre FormBodyWrapperFilter adds multiples of every multipart part to a request

Bumping RxJava to latest version

1.1.5 -> 1.1.10

fixes #1337

Was previously milliseconds/seconds when calculating whether to display the message about reporting failures.

fixes gh-126

* garciafl-1.1.x:
  Remove RibbonCommand suffix

This is a regression in HttpClientRibbonCommand

Fixes gh-1308

fixes gh-1285

 Clarify peer replication setup of eureka servers (Fixes #1251)

This fixes a case where the ribbon property is set `ribbon.IsSecure=false`, but because the eureka secure port was enabled, the uri created by ribbon was prefixed with https when it shouldn't have.

All places where isSecure is need have been consolidated into RibbonUtils.

fixes gh-1270

Previously with a `@FeignClient(name="http://bad_hostname")` would pass the start time test and fail later in runtime. The test now fails fast.

fixes gh-1257

Example uses the same value of eureka.client.serviceUrl.defaultZone for all peers. The eureka server knows who it is and doesn't attempt to replicate to itself [1], so this simplification will work.

[1]https://github.com/Netflix/eureka/blob/v1.4.10/eureka-core/src/main/java/com/netflix/eureka/registry/PeerAwareInstanceRegistryImpl.java#L616-L619

A problem in Spring Boot means that the extra conditions are combined
with OR not AND, so it's a bug to have more than one until that
gets fixed (in an AnyNestedCondition).

See https://github.com/spring-projects/spring-boot/pull/6672

What we're missing ATM is different documentation versions for different application versions. What this change does is that it's:

- finding out what is the current branch (e.g. 1.0.x)
- finding out out what is the name of the main adoc file (e.g. spring-cloud-sleuth)
- pulling the changes from gh-pages after checkout
- finding out what is the list of comma separated whitelisted branches (via the `docs.whitelisted.branches` prop)
- in gh-pages creating a folder with name of the branch  (e.g. /1.0.x)
copying all the docs/target/generated-docs/ to that folder
- if the branch from which we're calling the script is NOT master then we're changing the ${main.adoc}.html to index.html so that it's easier to access the docs (e.g. http://cloud.spring.io/spring-cloud-sleuth/1.0.x/)

Content-length not properly set when using the FormBodyWrapperFilter

Fix an issue where the FormBodyWrapperFilter would encode the form parameters differently than on the original request.

The FormBodyWrapperFilter handles the application/x-www-form-urlencoded. In the case of requests received by curl or javascript, the FormHttpMessageConverter will reencode the parameters differently (for example, '(' will be encoded while it's not with the javascript encodeURIComponent method).

While this doesn't create any problem, the content length was not properly set in HttpClientRibbonCommand. This causes the form params being stripped or the backend server would wait a long time for additional bytes depending on if the content length header was bigger/smaller than the actual data.