1. The 1500,1000 refers to the number of requests that can be executed concurrently. 
2. On an average, 900 - 200 parallel invocations per minute will be the maximum number of function invocations depending on the execution time of 1 - 5s respectively.
3. Yes, if you have 1 function with 3 endpoints and each endpoint receives 100 requests, the total number of requests will be 300. So, you will be charged 300*0.5 credits. 
4. Yes, if you have 3 functions with 1 endpoint and each function receives 100 requests, the total number of requests will be 300. So, you will be charged 300*0.5 credits. 

Note : The credits calculation for 3 and 4 has been done with a function memory of 256 mb and execution time lesser than 1 second. It can vary accordingly to the memory and execution time.