rewrite circleci config (#558)

* rewrite circleci config with better syntax

and try using grcov for coverage

* attempt to use less memory and to rerun if memory lack

* actually remove use of kcov

* use the right var

* try looping

* run test serialized

* use feature where we should and make macro path independant

* cargo fmt and remove some '"'

* add zip to dependancies

* reduce number of used thread after first failure

* push translations to crowdin on commit to master

* don't cache result of coverage

script/run_browser_test.sh

@@ -3,16 +3,12 @@ set -eo pipefail
 
 export ROCKET_SECRET_KEY="AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA="
 
-mkdir -p "target/cov/plume"
-mkdir -p "target/cov/plm"
-plm='kcov --exclude-pattern=/.cargo,/usr/lib --verify target/cov/plm plm'
+plm migration run
+plm migration redo
+plm instance new -d plume-test.local -n plume-test
+plm users new -n admin -N 'Admin' -e 'email@exemple.com' -p 'password'
 
-$plm migration run
-$plm migration redo
-$plm instance new -d plume-test.local -n plume-test
-$plm users new -n admin -N 'Admin' -e 'email@exemple.com' -p 'password'
-
-kcov --exclude-pattern=/.cargo,/usr/lib --verify target/cov/plume plume &
+plume &
 caddy -conf /Caddyfile &
 
 until curl http://localhost:7878/test/health -f; do sleep 1; done 2>/dev/null >/dev/null
@@ -22,4 +18,4 @@ python3 -m unittest *.py
 
 kill -SIGINT  %1
 kill -SIGKILL %2
-wait
+sleep 15