fix: add sanity checks to prepare_new_block

[sdbondi]

Description

• Add basic sanity checks to prepare_new_block
• Check calc_median_timestamp invariant holds when used in prepare_new_block
• get_new_block GRPC call returns invalid argument error if invalid arguments are passed to prepare_new_block
• Add new unit tests (tests previous panic)

Motivation and Context

Possible panic observed in cucumber test is prevented

Testing scenario "Verify all coinbases in hybrid mining are accounted for"
.seed: SEED_A
........[mmProxy client] Tip at 0 ...(stopping at 100)
stderr: thread 'tokio-runtime-worker' panicked at 'calc_median_timestamp: timestamps cannot be empty', base_layer/core/src/validation/helpers.rs:79:5

stderr: stack backtrace:

stderr:    0: std::panicking::begin_panic

stderr:    1: tari_core::validation::helpers::calc_median_timestamp
   2: tari_core::chain_storage::blockchain_database::BlockchainDatabase<B>::prepare_new_block
   3: tari_core::chain_storage::async_db::trace_log
   4: tokio::loom::std::unsafe_cell::UnsafeCell<T>::with_mut

stderr:    5: tokio::runtime::task::harness::Harness<T,S>::poll

stderr:    6: tokio::runtime::blocking::pool::Inner::run
note: Some details are omitted, run with `RUST_BACKTRACE=full` for a verbose backtrace.

[mmProxy client] Tip at 35 ...(stopping at 100)
[mmProxy client] Tip at 95 ...(stopping at 100)

How Has This Been Tested?

New unit tests + existing tests

[delta1]

Nice LGTM

[SWvheerden]

One of the sanity checks goes against a current config option/feature. We need to remove one of the two.

[SWvheerden]

LGTM