Document that the whole leading comment block is parsed as the runner JSON

[Thebeatkicks]

Every Intelligent Contract starts with a runner comment:

# { "Depends": "py-genlayer:<hash>" }

What is not documented anywhere we could find is that GenVM parses the entire
contiguous run of leading # lines as that one JSON document. So this is a
valid contract:

# { "Depends": "py-genlayer:1jb45aa8ynh2a9c9xn3b7qqh8sm5q93hwfp7jqmwsfhh8jpz09h6" }

# Anything you want to say about the contract.
from genlayer import *

and this one is rejected by the network, purely because the blank line is gone:

# { "Depends": "py-genlayer:1jb45aa8ynh2a9c9xn3b7qqh8sm5q93hwfp7jqmwsfhh8jpz09h6" }
# Anything you want to say about the contract.
from genlayer import *

A single # hello on line 2 is enough. Reproducible without spending anything,
against Bradbury (2026-09-04) — the first form returns a schema, the second
returns VMError: invalid_contract:

$ curl -s -X POST -H 'Content-Type: application/json' \
    -d "{\"jsonrpc\":\"2.0\",\"id\":1,\"method\":\"gen_getContractSchema\",\"params\":[{\"code\":\"$(base64 -w0 < contract.py)\"}]}" \
    https://rpc-bradbury.genlayer.com
{"jsonrpc":"2.0","result":null,"error":{"code":-32000,"message":"VMError: invalid_contract"},"id":1}

The execution trace names the cause, and the message is the confusing part:

error: { causes: [ 'trailing characters at line 1 column 84' ] }

Line 1 is 83 characters long and entirely correct. Column 84 is the newline —
the parser is reporting where line 2 got appended to it. On the attest call
against the resulting address the trace instead says
VMError(invalid_contract absent_runner_comment), which points at a runner
comment that is present and well-formed.

What makes this expensive is that nothing local catches it:

• genvm-lint check passes the broken file.
• The direct-mode test suite passes the broken file.
• genlayer deploy prints ✔ Contract deployed successfully. and returns a
  contract address.

The first real symptom is genlayer code <address> returning contract code not found at address …, or every call failing, against an address the CLI just
told you was deployed. We burned a deployment and an attest transaction
before finding it.

Suggested fixes, in order of how much they would have helped us:

1. Document the rule on the Intelligent Contracts introduction, next to the
   runner comment: the runner comment must be the only leading comment, or must
   be separated from any other comment by a blank line.
2. Have genvm-lint check it. It already parses the runner comment to
   resolve the version; rejecting a leading block that is not valid JSON would
   move this failure from a wasted deployment to a local lint error.
3. Improve the error. absent_runner_comment for a file whose first line
   is a runner comment sends readers to exactly the wrong place; naming the
   comment block would be enough.