perf: investigate whether `eth_estimateGas` can reuse the `EvmObserver` using `flush_and_report`

[Wodann]

Currently, the the eth_estimateGas constructs a new EvmObserver for each iteration of the binary search. Potentially, this is resulting in more deallocations and allocations, impacting performance.

In particular, this would be the case in the JSON-RPC provider if:

• tracing is enabled
• eth_estimateGas is called often

The simplest way to reproduce this would be to add a criterion benchmark that calls estimate_gas on the ProviderData repeatedly.

Action Items

• Add a benchmark that tests eth_estimateGas performance with & without tracing
• Validate whether reusing the same EvmObserver would improve performance.