Use Async.race in serverCLI start-all to fix initialisation bug

See merge request !404

This commit uses the `race` function from the async library to fix a bug
where exceptions raised from the server async wouldn't be caught by the
top-level code.

The bug was stemming from the fact that `runAllWorkers` is **blocking**,
despite its usage of `forConcurrently_`. Therefore we were never
actually running the `wait` function below, but rather we were hanging
waiting on the result of the first function.

As a result the server could die but the workers could keep the main
thread alive, causing the bug we just saw as part of #463.

Proper incremental TSV parser

See merge request !402

README: add info about system postgresql configuration

See merge request !403

This commit introduces/improves the `parseTvsWithDiagnostics`
function to parse the input TSV incrementally, collecting errors
as we go, and eventually reporting them upstream via the newly
added `emitTsvParseWarning` function.

This will pave the way for a more thorough refactoring to
be able to emit diagnostics after parsing.

Resolve "Import/export in SQLite format"

See merge request !393

Also, no need for `liftBase` in `getCorpusSQLite`.

Resolve "Implement temporary file storage"

See merge request !392

In an attempt to make the TSV parser more lenient for the absence of the
forms column, I have used NoHeader in the `decodeWithP` call, but that
caused the first line of the TSVs with a header to be interpreted as
terms, which is not what we want.

We would like to use IPC communication during testing, so that we
can leverage indpendent test runs. However, the implementation turned
out to be buggy in the now-unmaintainted `nanomsg`.

Concurrency is a good way to trigger any interleaving bug.

They can be enabled with `GGTX_LOG_LEVEL` during tests.

The bug is elsewhere.

This commit extends the timeout around the `readTChan` in the ping-pong
test to 60 seconds, mostly to account for CI slowness and to rule-out
potential source of flakyness.

The calls to `threadDelay` have been removed, as they doesn't seem to
affect the overall stability of the test (at least locally) -- the
point of the exercise is trying to expose as much as possible potential
problems which might be masked by active waits.

Relax a bit moveChecks to account for shared nodes

Closes #455

See merge request !398