ethcore/test_helpers: fix deadlock caused by double read lock

[BurtonQin]

Similar to #11766,

1. In block_hash(), the two numbers.read() may be interleaved by numbers.write() from another thread.
   https://github.com/openethereum/openethereum/blob/f8cbdadfaaef30febbeab1ec5b7331a74ec7856a/ethcore/src/test_helpers/test_client.rs#L360
   The fix is to use one numbers.read() instead of two.
2. In chain_info(), the two difficulty.read()(first_block.read(), ancient_block.read()) may be interleaved by their corresponding write lock from another thread.
   https://github.com/openethereum/openethereum/blob/f8cbdadfaaef30febbeab1ec5b7331a74ec7856a/ethcore/src/test_helpers/test_client.rs#L491-L502
   I lift the read locks and order them as they are declared. Currently, no conflicting lock order is found.

[sorpaas]

Nice! I remember we deliberately tried to keep lock as short as possible, but turned out in some situations they must be obtained in advance.