IO::Handle.readchars incorrectly assumes all chars are 2 bytes long

[p6rt]

Migrated from rt.perl.org#130840 (status was 'new')

Searchable as RT130840$

[p6rt]

From @zoffixznet

Based on the code, seems it's a naive kludge for the unimplemented nqp​::readcharsfh op. As a result, .readchars on JVM reads incorrect number
of chars, at times producing wrong data altogether​:

<ZoffixW> j​: "/tmp/Foo.pm6".IO.spurt​: "I♥u so very much"; "/tmp/Foo.pm6".IO.open.readchars(3).say
<camelia> rakudo-jvm fb4f16​: OUTPUT​: «I♥u ␤»
<ZoffixW> j​: "/tmp/Foo.pm6".IO.spurt​: "I love you so very much"; "/tmp/Foo.pm6".IO.open.readchars(3).say
<camelia> rakudo-jvm fb4f16​: OUTPUT​: «I love␤»
<ZoffixW> j​: "/tmp/Foo.pm6".IO.spurt​: "♥"; "/tmp/Foo.pm6".IO.open.readchars(1).say
<camelia> rakudo-jvm fb4f16​: OUTPUT​: «�␤»

[usev6]

I'm of the opinion that this is caused by nqp::chars working on Java's UTF-16 chars, not on graphemes: Raku/nqp#783.

Just mentioning this here to link the two issues.