Currently in this module we have a distinction between an "index" which
is a number between [0, 2^31 - 1] which indexes into the set of normal
or hardened child numbers. We also have a child *number*, which within
the library is an opaque type, but can be freely converted into/out of a
u32 in the consensus encoding which uses the MSB as a normal/hardened
flag and the other bits to encode the index.
Probably we want to change ChildNumber::From<u32> to some sort of
from_consensus_u32 method, but that's out of scope for this PR. What is
*in scope*, though is fixing the error types.
In the existing code we have three problems:
* Our error type for a bad index is called InvalidChildNumber, rather
than InvalidIndex, and the error message reflects this, which is wrong
and confusing. (Some with InvalidChildNumberFormat.)
* The InvalidChildNumberFormat is always constructed from a ParseIntError
from stdlib, but we always throw that away rather than preserving it.
* These two error variants only appear when parsing child numbers, or
derivation paths which are lists of child numbers, but they are part
of the main error enum.
a66ad97fb6