Reduce memory and s3 costs

[devinmatte]

Motivation

Max memory used by NTT is around 192mb, so we don't need to be paying for 256. Reducing to 220 so we have some headroom still without going too far over

https://app.datadoghq.com/serverless/aws/lambda?fromUser=false&group=&panel_end=1767631532770&panel_paused=false&panel_start=1765039532770&panel_tab=metrics&sp=%5B%7B%22p%22%3A%7B%22entityId%22%3A%22aws-lambda-functions%2Bntt-apihandler-hxdxktd0wizd%2Bus-east-1%2B473352343756%22%7D%2C%22i%22%3A%22lambda-panel%22%7D%5D&start=1767627360000&end=1767630960000&paused=false

We also are currently uploading empty s3 files for lines with 0 new trains or 0 sightings of new trains. We can stop doing that

Changes

• Drops memory demand to 220 from 256
• Checks if we have seen new trains before uploading a new train file to s3 (reducing s3 calls)
• Adds a type hint because I noticed it

Testing Instructions