Implement Station Look-Up

To actually route in a public transportation network, we require the translation from location / query to stations.

I see a few different scenarios that we might want to support.

# Name To Station

## Similarity measures

Entering the name of a station, or parts of it, we should be able to find the best match in the data.

For any string `s` and station `S`, we need a function that determines the likelihood of `s` referring to `S`. We could use different synonyms (e.g. `S+U Stationname` and `Stationname`) to refer to the same station.

Similarity could further use techniques like stemming, phonetics matching or similar to distinguish between possible candidates.

A possible hint to further strengthen the look-up could be the distance to the GPS position.

## Selecting stations

User enters name 'n'. If name 'n' can be matched with a high enough certainty, we directly select the station. If the likelihood of many stations is high enough, we should output a sorted list of possible candidates.

### Possible Features

By supporting multiple likely results (staring at maybe like 2/3 characters), we could implement autocomplete.

# Location to Station

For the look-up location to station we would need integration with walking durations. What I imagine is a preselection phase to select a Pareto-set of possible connections. For every line we can select the nearest (maybe two nearest/maybe per direction?) stations.
Within a maximal walking radius, we can pre-filter a set of candidates.

These candidates then need to be checked for their walking distances. This could be done with a walking table query to OSRM/mapbox-directions.

This gives a set of initial data. Next to this initial data, we also require a check for the direct connection (if it is close enough) in walking.

After a search for the best connections, we can then augment our response by querying the walk-api for an actual set of walking instructions to the stations/from the stations.

/ cc @daniel-j-h

The issue names no files, tests, or entry points. Start by mapping the existing GTFS station and routing flow, then review the proposed OSRM/mapbox-directions and walk-api integrations. Before implementation, define the matching, candidate selection, walking-radius, and autocomplete acceptance criteria.

Written by the indexing model from the issue text.