[ADP-3414] Optimize benchmark history tracking (#4755)

This  PR refactors and improve buildkite lib.

- Introduces a LimitsLock object to avoid `429` when the 200 API calls
per minute are surpassed.
- Factor out code from the benchmark-history exe into the buildkite lib

I also compute the benchmark history incrementally, by adding new data
points to the last checkpoint

- Exposes the csv database of benchmarks for further reuse
- Catch that csv while collecting new benchmark results, stopping the
collection there and simply amending it

.buildkite/pipeline.yml

@@ -466,6 +466,7 @@ steps:
         - trigger-benchmarks
       artifact_paths:
         - ./benchmark-history*.tgz
+        - ./benchmark-history.csv
       command: |
         ./scripts/buildkite/main/benchmark-history.sh
       agents:

.gitignore

@@ -88,3 +88,6 @@ test/e2e/.direnv/flake-profile*
 node_modules
 package-lock.json
 package.json
+
+# buildkite artifacts testing
+artifacts

lib/benchmarks/cardano-wallet-benchmarks.cabal

@@ -39,6 +39,7 @@ library
   hs-source-dirs:  src
   build-depends:
     , aeson
+    , attoparsec
     , base
     , bytestring
     , cardano-wallet
@@ -83,6 +84,7 @@ library
     , time
     , unliftio
     , unliftio-core
+    , unordered-containers
     , vector
     , wai-middleware-logging
     , with-utf8
@@ -275,6 +277,7 @@ executable benchmark-history
     , cardano-wallet-buildkite
     , cassava
     , containers
+    , contra-tracer
     , directory
     , filepath
     , exceptions