'#'' is not in the set qchar indirectly defined in BIP 21, and therefore
should be escaped.

[BIP 21](https://github.com/bitcoin/bips/blob/master/bip-0021.mediawiki#abnf-grammar):

>     labelparam     = "label=" *qchar
>     messageparam   = "message=" *qchar
>     otherparam     = qchar *qchar [ "=" *qchar ]
...
> Here, "qchar" corresponds to valid characters of an RFC 3986 URI query
> component, excluding the "=" and "&" characters, which this BIP takes
> as separators.

[RFC 3986 § 3.4](https://www.rfc-editor.org/rfc/rfc3986#section-3.4):

> The query component is indicated by the first question mark ("?")
> character and terminated by a number sign ("#") character or by the
> end of the URI.

[RFC 3986 Appendix A](https://www.rfc-editor.org/rfc/rfc3986#appendix-A):

>     pchar         = unreserved / pct-encoded / sub-delims / ":" / "@"
>     query         = *( pchar / "/" / "?" )
...
>     pct-encoded   = "%" HEXDIG HEXDIG
>     unreserved    = ALPHA / DIGIT / "-" / "." / "_" / "~"
...
>     sub-delims    = "!" / "$" / "&" / "'" / "(" / ")"
>                   / "*" / "+" / "," / ";" / "="

Although behavior for when encountering RFC 3986 fragments in BIP 21
URIs is not specified, according to RFC 3986 it is unambiguously not
query data and therefore should be excluded from BIP 21 query
parameters.

Fix escaping of '#' in parameter values and handling of unescaped '#' in uri

fix: unicode on exactly index 7 should not panic

is_some_and is stable only in Rust 1.70.0 and greater.

map_or(false, |s| s.eq_ignore_ascii_case(SCHEME)):
  - This method is used to handle the Option returned by get().
  - If get() returns None (i.e., the string is too short), map_or()
    returns false.
  - If get() returns Some(s), it applies the closure
    |s| s.eq_ignore_ascii_case(SCHEME).
     - This closure checks if the substring s is equal to SCHEME,
       ignoring ASCII case differences.
     - If they are equal, it returns true; otherwise, it returns
       false.

Replace is_some_and with map_or for MSRV

The error is only passed up via `source` when `std::error::Error` is
available. When the feature is off, it's dead code.

The code isn't actually dead, and is used in the bytes() function, but the
compiler's dead code analysis doesn't fully track usage through feature flags
and doesn't count if the pattern isn't considered a read of the field.