Pass non-segmental tiers on in layer-based reconstruction

[LinguList]

Non-segmental tiers are lost now, we should consider to pass them on for now, to allow for their inclusion in layer-based sound change. This can be trivially handled, but must be taken with caution. Tiers other than segments cannot be changed so far.

So an alternative may be to test syllable boundaries in sound change (given that we do not model sound change across whole multi-morphemic words).

[LinguList]

We can use standard stress markers and the dot as syllable boundary marker and include a @stress function.

[LinguList]

Stress has now been handled. I consider this as almost satisfying enough. We say: tiers can be used will be used, but typically, one should be able to compute them from the input sequence, avoiding user-defined custom tiers.