Assume I have a XML doc and a XSD schema file which defines mandatory
and optional element fields.

How can I now go recursively through the whole XML document

and find all elements which are optional (e.g. minoccurs=0)

That information is in XSD. I might be wrong, but I don't think it is
exposed in the XPath 2.0 Data Model with XSD annotations enabled. You
might get a list of the optional elements by (1) reading the schema,
(2) grepping the schema, or processing Saxon's normalized XSD
representation.

but currently EMPTY.

"Empty" might need a better definition. I'll take it to mean the absence
of nodes on the child axis, without taking the attribute axis into
account.

These empty, optional elements should be stripped off so that the
resulting XML does not contain them any more.

<xsl:stylesheet version="1.0"
  xmlns:xsl="http://www.w3.org/1999/XSL/Transform">

  <xsl:template match="eins[not(node())]"/>
  <xsl:template match="zwei[not(node())]"/>
  <xsl:template match="drei[not(node())]"/>

  <xsl:template match="@*|node()"><!-- identity template -->
    <xsl:copy><xsl:apply-templates select="@*|node()"/></xsl:copy>
  </xsl:template>