Update variable detection in UriComponentsBuilder#encode

This commit better aligns how URI variable placeholders are detected
in UriComponentsBuilder#encode (i.e. the pre-encoding of the literal
parts of a URI template) and how they are expanded later on.
The latter relies on a pattern that stops at the first closing '}'
which excludes the possibility for well-formed, nested placeholders
other than variables with regex syntax, e.g. "{year:\d{1,4}}".

UriComponentsBuilder#encode now also stops at the first closing '}' and
further ensures the placeholder is not empty and that it has '{' before
deciding to treat it as a URI variable.

Closes gh-26466

src/docs/asciidoc/web/web-uris.adoc

@@ -214,8 +214,10 @@ TIP: Consider ";", which is legal in a path but has reserved meaning. The first
 replaces ";", since it is a legal character in a path.
 
 For most cases, the first option is likely to give the expected result, because it treats URI
-variables as opaque data to be fully encoded, while option 2 is useful only if
-URI variables intentionally contain reserved characters.
+variables as opaque data to be fully encoded, while the second option is useful if URI
+variables do intentionally contain reserved characters. The second option is also useful
+when not expanding URI variables at all since that will also encode anything that
+incidentally looks like a URI variable.
 
 The following example uses the first option: