feat(runner): add tx-firehose standalone tx tool

[mkoura]

Add tx-firehose next to the existing tx-centrifuge in the regression runner, mirroring the integration in ../cardonnay/flake.nix (cardano-node ref leios-prototype). Both live on their own cardano-node refs and are built on demand, tx-centrifuge when ENABLE_TX_CENTRIFUGE is set and tx-firehose when ENABLE_TX_FIREHOSE is set.

The build logic is the same for every such tool, with the tool name as the only variable, so runner/source_tx_centrifuge.sh is generalized into runner/source_tx_tool.sh (tx_tool_build <tool> <rev> and tx_tool_print_path_prepend <tool>) and driven from a <tool>:<enable var>:<rev var>:<default rev> table in regression.sh.

That table is the single source of truth. The loop records the enabled tools in ENABLED_TX_TOOLS, which is exported and expanded into _req_cmds in the testrun step, so the availability check does not repeat the list of enable variables and adding a tool means editing one place. The list is space-separated, so tool names are rejected if they contain whitespace.

Neither tool belongs in the dev environment: having them in flake.nix would always build extra cardano-node instances from their refs just to enter the dev shell. They are therefore available only in a real testrun through regression.sh, and both are removed from
scripts/build_and_link_node_bins.sh and scripts/check_dev_env.sh (where tx-centrifuge would otherwise always be reported as missing). The now-unused common.sh sourcing in build_and_link_node_bins.sh goes away with them.

Dropping the extra flake input also re-resolved the remaining ones, so flake.lock bumps.