firestore: handle transient connection errors

[Kavantix]

Client

Firestore

Environment

Cloud Run

Go Environment
n/a

Expected behavior

Reading any document from cloud run should always succeed without connection issues, general firebase admin sdk documentation states it should handle any connection resets

Actual behavior

The syscall.ECONNRESET and syscall.ECONNREFUSED errors are not handled but instead passed as error to the caller.
This gives a message like:

rpc error: code = Unavailable desc = error reading from server: read tcp [fddf:3978:feb1:d745::c001]:55980->[2a00:1450:4001:81c::200a]:443: read: connection reset by peer

Notes

• This was recently fixed for the storage package: storage: Why "ECONNRESET on read" is not transient error  #9478 but should probably be done for all packages
• The storage package implements it using string comparison, but according to the go docs the error can be checked using errors.Is(err, syscall.ECONNRESET)