Profile and optimize OpenPedCan-api API server performance

[logstar]

OpenPedCan-api API is expected to respond 5 concurrent requests within 10 seconds, as suggested by @taylordm .

OpenPedCan-api API DEV site response times are summarized in the following boxplot. The median response time of /tpm/gene-all-cancer-gtex/plot endpoint is about 5 seconds per request, which will be longer after including more samples.

OpenPedCan-api API DEV/QA/PRD sites have multiple OpenPedCan-api API HTTP servers running on the backend, so multiple HTTP API requests can be handled concurrently.

OpenPedCan-api API DEV/QA/PRD sites need to be profiled with the following steps:

• Check the response time of 5 concurrent /tpm/gene-all-cancer-gtex/plot requests. Related to Load testing OpenPedcan-api server #27.
• Profile file transferring time.
• Profile runtime of R server code. This probably can be done using xrprof https://github.com/atheriel/xrprof.

Based on the profiling results, investigate and discuss:

• Does the API need to be optimized?
• Can the API be optimized to reach performance goals?
• How to optimize the API?

Test and implement promising optimization approaches.

cc @taylordm @chinwallaa @blackdenc