bug: table session pool gives up too fast on obtaining new session

[zveznicht]

Bug Report

YDB GO SDK version:

3.86.1

Issue:

On environments with small requests rate we occasionally get this error when running queries (table service):

failed to get item from pool after 100 attempts and 1.922558ms, pool has 1 items (0 busy, 1 idle, 0 create_in_progress): %!w(<nil>) at `github.com/ydb-platform/ydb-go-sdk/v3/internal/pool.(*Pool).getItem(pool.go:654)`

It seems it happens after 20-30 minutes without any request, which is larger than default session idle timeout. Judging by ydb sdk metrics it removes (almost?) all sessions from the pool (because they are all too old?), which is fine. What is not fine is that it doesn't seem to to try to create new session, certainty not 100 times or it doesn't wait for it to become available.