(avro) Support `StreamReadConstraints.maxDocumentLength` and `maxTokenCount` in Avro parser

Two `StreamReadConstraints` settings are ignored by the Avro backend:

**`maxTokenCount`**: `AvroParserImpl.nextToken()` (and `nextNameMatch()`) assign `_currToken` directly instead of going through `ParserMinimalBase._updateToken()` / `_nullSafeUpdateToken()`, so tokens are never counted — `currentTokenCount()` stays 0 and the limit is never enforced. Only the `nextName()` variants count.

**`maxDocumentLength`**: no `validateDocumentLength()` calls anywhere — neither in `JacksonAvroParserImpl`'s buffer reloads, nor up front for `byte[]` input in `AvroFactory`, nor for the Apache-decoder path (`ApacheAvroParserImpl`), which reads from the `InputStream` directly.

CBOR and Smile support both (see #649, #651).