Support for different sides of PATTERN

Hi,

There is a nice [feature](https://github.com/apertium/lexd/blob/main/Usage.md#alignment) for modeling split morphemes. However it is only applicable for LEXICON calls in the PATTERN section. Is it possible to extend this behavior to the PATTERN sections?

Here is my example:

```
PATTERNS
:Stem Stem

PATTERN Stem
Root GenderMarker

LEXICON Root
живжив:жи

LEXICON GenderMarker
:в
:й
```

I expect it to produce the following:

```
живжив:живжив
живжив:жийжий
```

Cheers!

Start with the alignment section in Usage.md and trace how split morphemes are handled for LEXICON calls. Then examine processing of PATTERN sections and compare the result with the provided PATTERNS, PATTERN, and LEXICON example. Done means the example produces both expected outputs, including the gender-marker variants.

Written by the indexing model from the issue text.