fixing the CI + workflow for deploying contracts to testnet

[konnov]

I did not expect that it will take me so much time (like a week) to fix the CI with all the recent Soroban changes. What we have here:

• An update to stellar/rs-soroban-sdk 22.0.0-rc.3.1 in the examples
• An update to @stellar/stellar-sdk v13.0
• A downgrade to rustc 1.81.0, otherwise, a few examples break on deployment with references-something-we-love-rust, see Rust 1.82+ enables reference-types / multivalue wasm target features stellar/rs-soroban-env#1469.
• The deployment scripts fixed and working with v22.0.0 on the testnet
• Calling __constructor instead of initialize in timelock, since cap-0058
• Since the Horizon API does not have result_meta_xdr any longer (two weeks earlier than announced), using getTransaction from the rpc.Server. Now we use both Horizon API and RPC API. Perhaps, we could switch to RPC completely in the future.
• Add workflow testnet.yml to deploy all the contracts we use, including the old version of xycloans. This workflow is only triggered on-demand or once per week, so we don't have to spam the testnet on every commit.

Optimally, it would be great to have a smaller PR, but all these things pulled one another.

[thpani]

Looks good, thank you for taking care of this! 🙏🏻

I just have a few small questions about small fragments that aren't clear: