Now we're in top-down aggregation order with our export again.

Removes some redundancy and helps separating the collection and the
serialization phases.

If given, render only the part of the document that has been identified
by the given `xml:id` to html.

code.

When preparing the TEI header of an exported corpus, the TEIHeaderStack
implementation asks TG-search for the root object's ancestor path(s) and
retrieves metadata for all of them. If we don't have read permission for
any of these objects, an AuthFaultException would have been thrown,
causing a 403. 

This fix causes these ancestors to be simply skipped (which will be
logged but not be noted in the corpus document itself).

It now asks for a document that I actually have access to.

Define the system property textgrid.sid to run.

When the service is set up, it will try to load configuration files:

1. internal file
2. /etc/textgrid/aggregator/aggregator.properties
3. ~/.aggregator.properties

The last one wins. The internal file will be filled with default values
according to the profiles selected at build time. A copy of the default
config will be generated at target/aggregator.properties; you can
generate it manually using the resources goal, e.g.:
mvn -Pesx1 resources:resources

Switched to version 7.9 of the TEI xslt stylesheets, pulled in via a git
submodule. Major overhaul of the EPUB export.

While our customized transformation works fine for single-text TEI
files, it generates invalid EPUB documents for corpus documents –
appearently the TOC generation that forms the base for OPF and NCX
metadata does not work correctly in these cases, and it is to complex to
fix just now.

The alternate approach for sourceTypes != XML now converts the corpus to
a single TEI document with a <tei:group> for every original TEI or
teiCorpus document, and runs a mostly unmodified original EPUB
transformation on this document. This leads to surplus headings, a
suboptimal TOC and some additional content, but at least these EPUBs
validate with epubcheck now.