Resolve "[Node Terms] On multiple Map terms, if the first term is already a map term, the count of the multiple map term is to zero"

Closes #471

See merge request !414

Ref #488

[BREAKING] refactoring of fc_url

Closes #465

See merge request !405

Forest of trees: restore hierarchical grouping of terms

Closes #313

See merge request !424

Generalise fully 'GeneralisedNgramsTree' and add more commentary.

[sqlite] improve performance of sqlite generation by proper PRAGMA

See merge request !431

Also, force WAL checkpoint at the end, otherwise the DB might be incomplete.

This commit generalises the API we have seen for Ngrams and
NgramsElement to Ngrams and NgramsRepoElement, making it suitable for
the NodeStory manipulations.

Furthermore, we use zippers to efficiently traverse forests.

This paves the way for a more disciplined way of handling forests of
ngrams.

Revert "Merge branch 'adinapoli/bump-process' into 'dev'"

See merge request !427

This reverts merge request !426

Update process library to 1.6.26.1

See merge request !426

It turns out that we don't need to call `pruneForest` inside
`searchtableNgrams`. After all, we are just displaying information that
should be correct in the database from the get-go, so why we would need
any of that?

It turns out that the fix for #313 to support duplicated ngrams during
importing broke hierarchical grouping.

I have just temporarily commented out the code and I need to find a fix,
better also writing a test to ensure that grouping will still work in the
future.

This function allows us to keep only the roots, and discard the
redundant children after building a 'Forest'.

I think we can use the NgramsTree and NgramsForest types to our
advantage for the 'searchTableNgrams' & co, so I have revamped that
module to put it to good use.

This commit changes the implementation of filterNgramsNodes to not
filter too aggressively nodes that are present in a path.

This causes a test failure that needs to be investigated.

Apparently, this is needed to run 'nix' inside tests (e.g. start corenlp)

Also, fix README since we merged tasty/hspec tests recently.

Take into account the async nature of createProcess. Add proper watch
so that we monitor that coreNLP is running.