eth1 deposit tracker cache populate flawed in slow networks

[dapplion]

Describe the bug

Ropsten is having very long block times. Our eth1 deposit tracker assumes constant times of 14 seconds, and doesn't populate the blocks necessary latter for block production.

DB cache has
blocks here
=========|      we are querying blocks here
    |    |    |xxxx|    |
day 0    1    2    3   now

Expected behavior

When computing which block to download too, it should be computed based on timestamp, not blockNumber. This way to cache will be populated up to where it's needed