Include failed transactions in raw datasets

[Theodus]

We currently filter out failed transactions from raw datasets. This behavior makes the data incomplete, and hinders verifiability (merkle roots in block headers cannot be re-computed from the extracted data). This is also a concern for users that would expect to have failed transaction data available for them to query, since it is recorded on-chain.

I think we have 2 options:

1. Include all transaction data in the existing tables. Rely on derived datasets to filter only successful transactions.
2. Add new tables for failed transaction data (transactions, logs, etc.)

My preference is option 1.

[fubhy]

Certainly #1. Failed transactions are part of the chain.

[abarmat]

Please option 1!

[leoyvens]

I suppose 1 is a better match to what people expect from using the JSON-RPC data model directly.

But we will need to change the type of value in the transactions table from Decimal128 to String.