Shell nix further steps

[pepeiborra]

I've done a bit more work on this, fixing compatibility with all supported ghc versions and adding some comments on the rationale behind the approach.

pepeiborra@devvm370 ~/s/ide (shell-nix)> nix-shell -run "cabal build --dry-run"
Resolving dependencies...
Build profile: -w ghc-8.8.3 -O1
In order, the following would be built (use -v for more details):
 - ghcide-0.2.0 (exe:ghcide-test-preprocessor) (first run)
 - lsp-test-0.11.0.2 (lib) (requires build)
 - streaming-commons-0.2.1.2 (lib) (requires download & build)
 - conduit-extra-1.3.5 (lib) (requires build)
 - hie-bios-0.6.1 (lib) (requires build)
 - ghcide-0.2.0 (lib) (first run)
 - haskell-language-server-0.2.2.0 (lib) (first run)
 - haskell-language-server-0.2.2.0 (exe:haskell-language-server-wrapper) (first run)
 - haskell-language-server-0.2.2.0 (exe:haskell-language-server) (first run)
 - haskell-language-server-0.2.2.0 (test:wrapper-test) (first run)
 - haskell-language-server-0.2.2.0 (test:func-test) (first run)

[fendor]

What do we need to update all this? Is niv update enough?
Will the build-plan be identical to non-nix builds?

[pepeiborra]

niv update only moves the version of Nixpkgs, but it doesn't change the set of Haskell packages used to create the environment.

The list of packages in the shell.nix does need to be maintained manually, and it's ok if it gets out of sync. The penalty in the latter case is less Nix cache reuse, or unnecessary packages in the environment, both of which are minor issues. There exist options to automate this, e.g. cabal2nix or haskell.nix, but the approach in this PR is imho the simplest and least opinionated one. To recap, the goal is not to replace Cabal with Nix, but to leverage the Nix cache.

I believe the Cabal build plan is identical to non-nix builds: Cabal first solves the build plan and then reuses whatever is available in the package database and builds everything else.

[fendor]

LGTM.
Thanks!