Inconsistent Charset error handling in native and JVM

[qwwdfsad]

JVM currently replaces all malformed input during decoding while Native silently ignores it.
underflow, overflow, malformed and unmappable flags should be consistent with JVM by default especially in the face of packed-based Input, where some error should be ignored until reading from the next packet of input, and some should be propagated or re-checked on next read, which potentially makes Charset usage error-prone

[fzhinkin]

We're rebooting the kotlinx-io development (see #131), all issues related to the previous versions will be closed. Consider reopening it the issue will remain (or the feature is still missing) in a new version.