Handle build_engine.exs failures

[mhanberg]

Description

When the engine fails to build due to dependency issues, the most likely cause is we updated a dependency in the engine and it needs to fetch new ones.

This often happens with local builds as the builds are scoped to the same version.

Solution

There are a few potential solutions

• alert the user they need to run expert engine clean.
• automatically retry and add force: true
• automatically delete the engine folder and retry.
• when making a local build, append a hash of mix.lock so that when deps change after a new local build, it rebuilds correctly

mhanberg: I think the last option is the best solution as this should only happen for local builds.

[doorgan]

This was weird, found it when testing #344:

13:48:41.880 [info] [Project file:///Users/dorgan/dev/rainet] Found elixir executable at /Users/dorgan/.asdf/shims/elixir
13:48:41.880 [debug] sent notification server -> client window/logMessage
13:48:41.880 [info] [Project file:///Users/dorgan/dev/rainet] Finding or building engine
13:48:41.880 [debug] sent notification server -> client window/logMessage
13:48:41.880 [debug] sent request server -> client window/workDoneProgress/create
13:48:41.880 [debug] sent notification server -> client $/progress
13:48:44.284 [debug] Building engine: * creating /Users/dorgan/.local/share/mise/installs/elixir/1.19.1/.mix/elixir/1-18-otp-28/rebar3
...
13:49:47.862 [info] Engine build available at: /Users/dorgan/Library/Application Support/Expert/0.1.0/elixir-1.18.4-erts-16.1.2/ff34f4c9e4410e4701b4feda78da3818/_build/dev_ns

• It's using the asdf shim (correct)
• It's putting rebar3 in the mise 1.19.1 (incorrect)
• It put the engine in the expert using elixir 1.18.4 (correct)

Everything works afterwards, but I wonder if the --if-missing in the build script or similar is causing hex/rebar to be installed in the wrong elixir/otp version directory and that's what causes the dependency issues