add initial capacity based Document constructor [LUCENE-2539]

When loading field from the index without a FieldSelector, and often in user code we know exactly how many fields a Document should have. This patch will simply add the ability to allocate Document memory for precisely.

I will include as a separate patch where this would be useful in conjunction with patch #3352.

---
Migrated from [LUCENE-2539](https://issues.apache.org/jira/browse/LUCENE-2539) by Woody Anderson, updated May 09 2016
Attachments: [LUCENE-2539.patch](https://apache.github.io/lucene-jira-archive/attachments/LUCENE-2539/LUCENE-2539.patch) (versions: 3)

Start by reviewing the Document constructor API and the attached LUCENE-2539.patch to understand the proposed entry point. Done means Document can be created with an initial field capacity and preserves existing construction behavior; verify the change against the project's relevant tests, which the issue does not name.

Written by the indexing model from the issue text.