The text formatting path used by SAMRecord.toString() / getSAMString() has been
cleaned up:

- SAMTextWriter.getSAMString no longer uses a JVM-wide synchronized static cache
  of a shared StringWriter/SAMTextWriter. It now allocates a fresh StringWriter
  per call. Without this change, every toString() call would have taken a global
  monitor.
- SAMTextWriter.writeAlignment is split into a private writeAlignmentNoNewline
  plus a thin wrapper that appends '\n'. getSAMString uses the no-newline
  variant, so the result no longer needs to be trimmed.

BREAKING CHANGE: SAMRecord.getSAMString() no longer terminates the returned
String with a newline character. This brings SAMRecord into line with the other
getSAMString() implementations (SAMSequenceRecord, SAMReadGroupRecord,
SAMProgramRecord, SAMFileHeader), which already return newline-free strings.
Callers that relied on the trailing '\n' as a separator (e.g. concatenating two
records' SAM strings) must now insert their own separator. Callers that stripped
the newline with .trim() can drop the call.

SAMRecord.format() was deprecated in favor of getSAMString() because it didn't
guarantee a valid SAM text representation. Now that getSAMString() is the
canonical formatter (and is used by toString()), drop format() and the private
helpers that supported it: addField, formatTagValue, safeEquals.

Also remove the SRALazyRecord.format() override, which existed only because
format() bypassed lazy attribute loading; getSAMString() goes through
SRALazyRecord.getBinaryAttributes() which already handles lazy loading. And
delete a couple of stale commented-out debug prints that referenced format().

BREAKING CHANGE: SAMRecord.format() has been removed. Callers should use
SAMRecord.getSAMString() instead.