Fix: Handle incoming labels with invalid UTF-8

[vancwo]

It's possible that incoming labels contain invalid UTF-8 characters. This results in a panic. This fix sanitizes the label's string to ensure only valid UTF-8 characters are included, by replacing invalid characters with � (REPLACEMENT CHARACTER)

[vancwo]

I encounted this bug while debugging a connection to Azure Postgres Flexible Server. It seems Azure's monitoring system sets the backend type to some strange characters:

This resulted in a panic during runtime

panic: label value "Єc!\xbap" is not valid UTF-8

goroutine 30 [running]:
github.com/prometheus/client_golang/prometheus.MustNewConstMetric(...)
        /go/pkg/mod/github.com/prometheus/[email protected]/prometheus/value.go:129
main.queryNamespaceMapping(0xc0000b8790, {0xafab8a, 0x10}, {{0xc000162900, 0x7, 0x8}, 0xc0001778c0, 0x1, 0x0})
        /app/cmd/postgres_exporter/namespace.go:161 +0x1b5b
main.queryNamespaceMappings(0xc000154700, 0xc0000b8790)
        /app/cmd/postgres_exporter/namespace.go:227 +0x5f2
main.(*Server).Scrape(0xc0000b8790, 0xc000154700, 0x90?)
        /app/cmd/postgres_exporter/server.go:125 +0x133
main.(*Exporter).scrapeDSN(0xc0001380c0, 0xc000154700, {0xc000166000, 0x7a})
        /app/cmd/postgres_exporter/datasource.go:114 +0x132
main.(*Exporter).scrape(0xc0001380c0, 0xc000154700)
        /app/cmd/postgres_exporter/postgres_exporter.go:681 +0x16b
main.(*Exporter).Collect(0xc0001380c0, 0xc000154700)
        /app/cmd/postgres_exporter/postgres_exporter.go:570 +0x25
github.com/prometheus/client_golang/prometheus.(*Registry).Gather.func1()
        /go/pkg/mod/github.com/prometheus/[email protected]/prometheus/registry.go:458 +0xe5
created by github.com/prometheus/client_golang/prometheus.(*Registry).Gather in goroutine 28
        /go/pkg/mod/github.com/prometheus/[email protected]/prometheus/registry.go:467 +0x568

I was able to make this fix and verify it works myself. If this pull request is not appropriate in its current state, I would appreciate it if a maintainer could take the time to fix this bug. Thanks!