feat: add cache to terrakit data fetch

[Beldine-Moturi]

Summary

Related Issue (optional)

How to test this PR?

• Update/redeploy the studio with changes in this branch

• Use this image in the pipelines deployment, specifically the terrakit-data-fetch pipeline pod: quay.io/geospatial-studio/geostudio-pipelines:cache-inference-images

• Submit sample inference request with bbox option.

• Check that /data/cache is created and sample files saved in it

• Submit second request with the exact same payload.

• Confirm that the second submission does not fetch data from terrakit in the terrakit-data-fetch pipeline step, but rather uses the saved files in the cache. (Sample logs for both requests attachaed below)

• Sample logs for new request:
  a1ab0320-65d0-49f6-abb7-c3c9a2bc5a16-task_0-terrakit-data-fetch-stdout (1).log

• Sample logs for duplicate request:
  b314a45b-d266-4561-8a9d-1c5c51f41c99-task_0-terrakit-data-fetch-stdout.log

Screenshots / Logs (optional)

Checklist

• This PR targets the main branch
• I have added or updated relevant docs.
• I have not included any secrets or credentials.
• Linting and formatting checks pass.

[fMurugi]

1. Can we have atomic writes.

In the case we have N pods writing or reading tp avoid partial files read:
Time →
[ create file ][ write... ][ write... ][ write... ]
↑
another pod reads here → 💥 corrupted/partial file

 **Sol:**
     Time →
  [ write tmp file completely ]
  [ atomically rename ]

Other pods:

• before rename → file doesn't exist
• after rename → file is 100% complete

2. The indexing should be done for each day ,so when we fetch data for 18-20 .we cache data for each day ,so as for the next request we have if dates expand or reduce, we can just fetch for the additional dates not all of the other dates.eg 18-21 will fetch only the missing date.